[[2305.05479] Energy-Efficient Mining for Blockchain-Enabled IoT Applications](http://arxiv.org/abs/2305.05479) #secureWhat are the optimal times for an Internet of Things (IoT) device to act as a blockchain miner? The aim is to minimize the energy consumed by low-power IoT devices that log their data into a secure (tamper-proof) distributed ledger. We formulate the energy-efficient blockchain mining for IoT devices as a multiple-stopping time partially observed Markov decision process (POMDP) to maximize the probability of adding a block in the blockchain; we also present a model to optimize the number of stops (mining instants). In general, POMDPs are computationally intractable to solve, but we show mathematically using submodularity that the optimal mining policy has a useful structure: 1) it is monotone in belief space, and 2) it exhibits a threshold structure, which divides the belief space into two connected sets. Exploiting the structural results, we formulate a computationally-efficient linear mining policy for the blockchain-enabled IoT device. We present a policy gradient technique to optimize the parameters of the linear mining policy. Finally, we use synthetic and real Bitcoin datasets to study the performance of our proposed mining policy. We demonstrate the energy efficiency achieved by the optimal linear mining policy in contrast to other heuristic strategies.
[[2305.05282] Fooling State-of-the-Art Deepfake Detection with High-Quality Deepfakes](http://arxiv.org/abs/2305.05282) #securityDue to the rising threat of deepfakes to security and privacy, it is most important to develop robust and reliable detectors. In this paper, we examine the need for high-quality samples in the training datasets of such detectors. Accordingly, we show that deepfake detectors proven to generalize well on multiple research datasets still struggle in real-world scenarios with well-crafted fakes. First, we propose a novel autoencoder for face swapping alongside an advanced face blending technique, which we utilize to generate 90 high-quality deepfakes. Second, we feed those fakes to a state-of-the-art detector, causing its performance to decrease drastically. Moreover, we fine-tune the detector on our fakes and demonstrate that they contain useful clues for the detection of manipulations. Overall, our results provide insights into the generalization of deepfake detectors and suggest that their training datasets should be complemented by high-quality fakes since training on mere research data is insufficient.
[[2305.05499] Effects of Real-Life Traffic Sign Alteration on YOLOv7- an Object Recognition Model](http://arxiv.org/abs/2305.05499) #securityThe advancement of Image Processing has led to the widespread use of Object Recognition (OR) models in various applications, such as airport security and mail sorting. These models have become essential in signifying the capabilities of AI and supporting vital services like national postal operations. However, the performance of OR models can be impeded by real-life scenarios, such as traffic sign alteration. Therefore, this research investigates the effects of altered traffic signs on the accuracy and performance of object recognition models. To this end, a publicly available dataset was used to create different types of traffic sign alterations, including changes to size, shape, color, visibility, and angles. The impact of these alterations on the YOLOv7 (You Only Look Once) model's detection and classification abilities were analyzed. It reveals that the accuracy of object detection models decreases significantly when exposed to modified traffic signs under unlikely conditions. This study highlights the significance of enhancing the robustness of object detection models in real-life scenarios and the need for further investigation in this area to improve their accuracy and reliability.
[[2305.05004] A Survey on the Geographic Diversity of Usable Privacy and Security Research](http://arxiv.org/abs/2305.05004) #securityIn human factor fields such as human-computer interaction (HCI), psychology, and behavioral sciences, researchers have been concerned that participant samples are skewed toward WEIRD, i.e., participants mostly come from Western, Educated, Industrialized, Rich, and Democratic societies. This WEIRD skew may affect the generalizability of study results and hinder understanding of diverse participant populations and their cultural differences. The usable privacy and security (UPS) field has inherited many research methodologies from research on human factor fields such as HCI. We conducted a literature review to understand the extent to which participant samples in UPS papers were WEIRD and the characteristics of the methodologies and research topics in each user study recruiting Western or non-Western participants. We found that the skew toward WEIRD in UPS is greater than that in HCI. Geographic and linguistic barriers in the study methods and recruitment methods may cause researchers to conduct a user study locally. In addition, many papers did not report participant demographics, which could hinder the replication of the reported studies, leading to low reproducibility. We provide the following suggestions to improve geographic diversity: facilitate replication studies, improve reproducibility, address issues of study and recruiting methods, diversify researchers, and facilitate research on the topics for non-WEIRD populations.
[[2305.05108] Socio-Technical Security Modelling: Analysis of State-of-the-Art, Application, and Maturity in Critical Industrial Infrastructure Environments/Domains](http://arxiv.org/abs/2305.05108) #securityThis study explores the state-of-the-art, application, and maturity of socio-technical security models for industries and sectors dependent on CI and investigates the gap between academic research and industry practices concerning the modelling of both the social and technical aspects of security. Systematic study and critical analysis of literature show that a steady and growing on socio-technical security M&S approaches is emerging, possibly prompted by the growing recognition that digital systems and workplaces do not only comprise technologies, but also social (human) and sometimes physical elements.
[[2305.05309] PSP Framework: A novel risk assessment method in compliance with ISO/SAE-21434](http://arxiv.org/abs/2305.05309) #securityAs more cars connect to the internet and other devices, the automotive market has become a lucrative target for cyberattacks. This has made the industry more vulnerable to security threats. As a result, car manufacturers and governments are working together to reduce risks and prevent cyberattacks in the automotive sector. However, existing attack feasibility models derived from the information technology field may not always provide accurate assessments of the potential risks faced by Vehicle Electronic Control Units in different operating conditions and domains. This paper introduces the PUNCH Softronix and Politecnico di Torino (PSP) framework to address this issue. This framework is designed to provide accurate assessments compatible with the attack feasibility models defined by the automotive product security standards. The PSP framework utilizes social sentiment analysis to evaluate the real threat risk levels.
[[2305.05343] Data Protection and Security Issues With Network Error Logging](http://arxiv.org/abs/2305.05343) #securityNetwork Error Logging helps web server operators detect operational problems in real-time to provide fast and reliable services. This paper analyses Network Error Logging from two angles. Firstly, this paper overviews Network Error Logging from the data protection view. The ePrivacy Directive requires consent for non-essential access to the end devices. Nevertheless, the Network Error Logging design does not allow limiting the tracking to consenting users. Other issues lay in GDPR requirements for transparency and the obligations in the contract between controllers and processors of personal data. Secondly, this paper explains Network Error Logging exploitations to deploy long-time trackers to the victim devices. Even though users should be able to disable Network Error Logging, it is not clear how to do so. Web server operators can mitigate the attack by configuring servers to preventively remove policies that adversaries might have added.
[[2305.05506] Balancing Privacy and Security in Federated Learning with FedGT: A Group Testing Framework](http://arxiv.org/abs/2305.05506) #securityWe propose FedGT, a novel framework for identifying malicious clients in federated learning with secure aggregation. Inspired by group testing, the framework leverages overlapping groups of clients to detect the presence of malicious clients in the groups and to identify them via a decoding operation. The identified clients are then removed from the training of the model, which is performed over the remaining clients. FedGT strikes a balance between privacy and security, allowing for improved identification capabilities while still preserving data privacy. Specifically, the server learns the aggregated model of the clients in each group. The effectiveness of FedGT is demonstrated through extensive experiments on the MNIST and CIFAR-10 datasets, showing its ability to identify malicious clients with low misdetection and false alarm probabilities, resulting in high model utility.
[[2305.05062] Indoor Localization and Multi-person Tracking Using Privacy Preserving Distributed Camera Network with Edge Computing](http://arxiv.org/abs/2305.05062) #privacyLocalization of individuals in a built environment is a growing research topic. Estimating the positions, face orientation (or gaze direction) and trajectories of people through space has many uses, such as in crowd management, security, and healthcare. In this work, we present an open-source, low-cost, scalable and privacy-preserving edge computing framework for multi-person localization, i.e. estimating the positions, orientations, and trajectories of multiple people in an indoor space. Our computing framework consists of 38 Tensor Processing Unit (TPU)-enabled edge computing camera systems placed in the ceiling of the indoor therapeutic space. The edge compute systems are connected to an on-premise fog server through a secure and private network. A multi-person detection algorithm and a pose estimation model run on the edge TPU in real-time to collect features which are used, instead of raw images, for downstream computations. This ensures the privacy of individuals in the space, reduces data transmission/storage and improves scalability. We implemented a Kalman filter-based multi-person tracking method and a state-of-the-art body orientation estimation method to determine the positions and facing orientations of multiple people simultaneously in the indoor space. For our study site with size of 18,000 square feet, our system demonstrated an average localization error of 1.41 meters, a multiple-object tracking accuracy score of 62%, and a mean absolute body orientation error of 29{\deg}, which is sufficient for understanding group activity behaviors in indoor environments. Additionally, our study provides practical guidance for deploying the proposed system by analyzing various elements of the camera installation with respect to tracking accuracy.
[[2305.05391] Privacy-preserving Adversarial Facial Features](http://arxiv.org/abs/2305.05391) #privacyFace recognition service providers protect face privacy by extracting compact and discriminative facial features (representations) from images, and storing the facial features for real-time recognition. However, such features can still be exploited to recover the appearance of the original face by building a reconstruction network. Although several privacy-preserving methods have been proposed, the enhancement of face privacy protection is at the expense of accuracy degradation. In this paper, we propose an adversarial features-based face privacy protection (AdvFace) approach to generate privacy-preserving adversarial features, which can disrupt the mapping from adversarial features to facial images to defend against reconstruction attacks. To this end, we design a shadow model which simulates the attackers' behavior to capture the mapping function from facial features to images and generate adversarial latent noise to disrupt the mapping. The adversarial features rather than the original features are stored in the server's database to prevent leaked features from exposing facial information. Moreover, the AdvFace requires no changes to the face recognition network and can be implemented as a privacy-enhancing plugin in deployed face recognition systems. Extensive experimental results demonstrate that AdvFace outperforms the state-of-the-art face privacy-preserving methods in defending against reconstruction attacks while maintaining face recognition accuracy.
[[2305.05602] Privacy-Preserving Collaborative Chinese Text Recognition with Federated Learning](http://arxiv.org/abs/2305.05602) #privacyIn Chinese text recognition, to compensate for the insufficient local data and improve the performance of local few-shot character recognition, it is often necessary for one organization to collect a large amount of data from similar organizations. However, due to the natural presence of private information in text data, different organizations are unwilling to share private data, such as addresses and phone numbers. Therefore, it becomes increasingly important to design a privacy-preserving collaborative training framework for the Chinese text recognition task. In this paper, we introduce personalized federated learning (pFL) into the Chinese text recognition task and propose the pFedCR algorithm, which significantly improves the model performance of each client (organization) without sharing private data. Specifically, based on CRNN, to handle the non-iid problem of client data, we add several attention layers to the model and design a two-stage training approach for the client. In addition, we fine-tune the output layer of the model using a virtual dataset on the server, mitigating the problem of character imbalance in Chinese documents. The proposed approach is validated on public benchmarks and two self-built real-world industrial scenario datasets. The experimental results show that the pFedCR algorithm can improve the performance of local personalized models while also improving their generalization performance on other client data domains. Compared to local training within an organization, pFedCR improves model performance by about 20%. Compared to other state-of-the-art personalized federated learning methods, pFedCR improves performance by 6%~8%. Moreover, through federated learning, pFedCR can correct erroneous information in the ground truth.
[[2305.05247] Leveraging Generative AI Models for Synthetic Data Generation in Healthcare: Balancing Research and Privacy](http://arxiv.org/abs/2305.05247) #privacyThe widespread adoption of electronic health records and digital healthcare data has created a demand for data-driven insights to enhance patient outcomes, diagnostics, and treatments. However, using real patient data presents privacy and regulatory challenges, including compliance with HIPAA and GDPR. Synthetic data generation, using generative AI models like GANs and VAEs offers a promising solution to balance valuable data access and patient privacy protection. In this paper, we examine generative AI models for creating realistic, anonymized patient data for research and training, explore synthetic data applications in healthcare, and discuss its benefits, challenges, and future research directions. Synthetic data has the potential to revolutionize healthcare by providing anonymized patient data while preserving privacy and enabling versatile applications.
[[2305.05355] Turning Privacy-preserving Mechanisms against Federated Learning](http://arxiv.org/abs/2305.05355) #privacyRecently, researchers have successfully employed Graph Neural Networks (GNNs) to build enhanced recommender systems due to their capability to learn patterns from the interaction between involved entities. In addition, previous studies have investigated federated learning as the main solution to enable a native privacy-preserving mechanism for the construction of global GNN models without collecting sensitive data into a single computation unit. Still, privacy issues may arise as the analysis of local model updates produced by the federated clients can return information related to sensitive local data. For this reason, experts proposed solutions that combine federated learning with Differential Privacy strategies and community-driven approaches, which involve combining data from neighbor clients to make the individual local updates less dependent on local sensitive data. In this paper, we identify a crucial security flaw in such a configuration, and we design an attack capable of deceiving state-of-the-art defenses for federated learning. The proposed attack includes two operating modes, the first one focusing on convergence inhibition (Adversarial Mode), and the second one aiming at building a deceptive rating injection on the global federated model (Backdoor Mode). The experimental results show the effectiveness of our attack in both its modes, returning on average 60% performance detriment in all the tests on Adversarial Mode and fully effective backdoors in 93% of cases for the tests performed on Backdoor Mode.
[[2305.05404] Probabilistic Detection of GNSS Spoofing using Opportunistic Information](http://arxiv.org/abs/2305.05404) #protectGlobal Navigation Satellite Systems (GNSS) are integrated into many devices. However, civilian GNSS signals are usually not cryptographically protected. This makes attacks that forge signals relatively easy. Considering modern devices often have network connections and onboard sensors, the proposed here Probabilistic Detection of GNSS Spoofing (PDS) scheme is based on such opportunistic information. PDS has at its core two parts. First, a regression problem with motion model constraints, which equalizes the noise of all locations considering the motion model of the device. Second, a Gaussian process, that analyzes statistical properties of location data to construct uncertainty. Then, a likelihood function, that fuses the two parts, as a basis for a Neyman-Pearson lemma (NPL)-based detection strategy. Our experimental evaluation shows a performance gain over the state-of-the-art, in terms of attack detection effectiveness.
[[2305.05133] Generating Phishing Attacks using ChatGPT](http://arxiv.org/abs/2305.05133) #attackThe ability of ChatGPT to generate human-like responses and understand context has made it a popular tool for conversational agents, content creation, data analysis, and research and innovation. However, its effectiveness and ease of accessibility makes it a prime target for generating malicious content, such as phishing attacks, that can put users at risk. In this work, we identify several malicious prompts that can be provided to ChatGPT to generate functional phishing websites. Through an iterative approach, we find that these phishing websites can be made to imitate popular brands and emulate several evasive tactics that have been known to avoid detection by anti-phishing entities. These attacks can be generated using vanilla ChatGPT without the need of any prior adversarial exploits (jailbreaking).
[[2305.05253] Attack Named Entity Recognition by Entity Boundary Interference](http://arxiv.org/abs/2305.05253) #attackNamed Entity Recognition (NER) is a cornerstone NLP task while its robustness has been given little attention. This paper rethinks the principles of NER attacks derived from sentence classification, as they can easily violate the label consistency between the original and adversarial NER examples. This is due to the fine-grained nature of NER, as even minor word changes in the sentence can result in the emergence or mutation of any entities, resulting in invalid adversarial examples. To this end, we propose a novel one-word modification NER attack based on a key insight, NER models are always vulnerable to the boundary position of an entity to make their decision. We thus strategically insert a new boundary into the sentence and trigger the Entity Boundary Interference that the victim model makes the wrong prediction either on this boundary word or on other words in the sentence. We call this attack Virtual Boundary Attack (ViBA), which is shown to be remarkably effective when attacking both English and Chinese models with a 70%-90% attack success rate on state-of-the-art language models (e.g. RoBERTa, DeBERTa) and also significantly faster than previous methods.
[[2305.05070] Distributed Detection over Blockchain-aided Internet of Things in the Presence of Attacks](http://arxiv.org/abs/2305.05070) #attackDistributed detection over a blockchain-aided Internet of Things (BIoT) network in the presence of attacks is considered, where the integrated blockchain is employed to secure data exchanges over the BIoT as well as data storage at the agents of the BIoT. We consider a general adversary model where attackers jointly exploit the vulnerability of IoT devices and that of the blockchain employed in the BIoT. The optimal attacking strategy which minimizes the Kullback-Leibler divergence is pursued. It can be shown that this optimization problem is nonconvex, and hence it is generally intractable to find the globally optimal solution to such a problem. To overcome this issue, we first propose a relaxation method that can convert the original nonconvex optimization problem into a convex optimization problem, and then the analytic expression for the optimal solution to the relaxed convex optimization problem is derived. The optimal value of the relaxed convex optimization problem provides a detection performance guarantee for the BIoT in the presence of attacks. In addition, we develop a coordinate descent algorithm which is based on a capped water-filling method to solve the relaxed convex optimization problem, and moreover, we show that the convergence of the proposed coordinate descent algorithm can be guaranteed.
[[2305.05057] Crack Detection of Asphalt Concrete Using Combined Fracture Mechanics and Digital Image Correlation](http://arxiv.org/abs/2305.05057) #robustCracking is a common failure mode in asphalt concrete (AC) pavements. Many tests have been developed to characterize the fracture behavior of AC. Accurate crack detection during testing is crucial to describe AC fracture behavior. This paper proposed a framework to detect surface cracks in AC specimens using two-dimensional digital image correlation (DIC). Two significant drawbacks in previous research in this field were addressed. First, a multi-seed incremental reliability-guided DIC was proposed to solve the decorrelation issue due to large deformation and discontinuities. The method was validated using synthetic deformed images. A correctly implemented analysis could accurately measure strains up to 450\%, even with significant discontinuities (cracks) present in the deformed image. Second, a robust method was developed to detect cracks based on displacement fields. The proposed method uses critical crack tip opening displacement ($\delta_c$) to define the onset of cleavage fracture. The proposed method relies on well-developed fracture mechanics theory. The proposed threshold $\delta_c$ has a physical meaning and can be easily determined from DIC measurement. The method was validated using an extended finite element model. The framework was implemented to measure the crack propagation rate while conducting the Illinois-flexibility index test on two AC mixes. The calculated rates could distinguish mixes based on their cracking potential. The proposed framework could be applied to characterize AC cracking phenomenon, evaluate its fracture properties, assess asphalt mixture testing protocols, and develop theoretical models.
[[2305.05095] Less is More: Removing Text-regions Improves CLIP Training Efficiency and Robustness](http://arxiv.org/abs/2305.05095) #robustThe CLIP (Contrastive Language-Image Pre-training) model and its variants are becoming the de facto backbone in many applications. However, training a CLIP model from hundreds of millions of image-text pairs can be prohibitively expensive. Furthermore, the conventional CLIP model doesn't differentiate between the visual semantics and meaning of text regions embedded in images. This can lead to non-robustness when the text in the embedded region doesn't match the image's visual appearance. In this paper, we discuss two effective approaches to improve the efficiency and robustness of CLIP training: (1) augmenting the training dataset while maintaining the same number of optimization steps, and (2) filtering out samples that contain text regions in the image. By doing so, we significantly improve the classification and retrieval accuracy on public benchmarks like ImageNet and CoCo. Filtering out images with text regions also protects the model from typographic attacks. To verify this, we build a new dataset named ImageNet with Adversarial Text Regions (ImageNet-Attr). Our filter-based CLIP model demonstrates a top-1 accuracy of 68.78\%, outperforming previous models whose accuracy was all below 50\%.
[[2305.05222] FishRecGAN: An End to End GAN Based Network for Fisheye Rectification and Calibration](http://arxiv.org/abs/2305.05222) #robustWe propose an end-to-end deep learning approach to rectify fisheye images and simultaneously calibrate camera intrinsic and distortion parameters. Our method consists of two parts: a Quick Image Rectification Module developed with a Pix2Pix GAN and Wasserstein GAN (W-Pix2PixGAN), and a Calibration Module with a CNN architecture. Our Quick Rectification Network performs robust rectification with good resolution, making it suitable for constant calibration in camera-based surveillance equipment. To achieve high-quality calibration, we use the straightened output from the Quick Rectification Module as a guidance-like semantic feature map for the Calibration Module to learn the geometric relationship between the straightened feature and the distorted feature. We train and validate our method with a large synthesized dataset labeled with well-simulated parameters applied to a perspective image dataset. Our solution has achieved robust performance in high-resolution with a significant PSNR value of 22.343.
[[2305.05400] Investigating the Corruption Robustness of Image Classifiers with Random Lp-norm Corruptions](http://arxiv.org/abs/2305.05400) #robustRobustness is a fundamental property of machine learning classifiers to achieve safety and reliability. In the fields of adversarial robustness and formal robustness verification of image classification models, robustness is commonly defined as the stability to all input variations within an Lp-norm distance. However, robustness to random corruptions is usually improved and evaluated using variations observed in the real-world, while mathematically defined Lp-norm corruptions are rarely considered. This study investigates the use of random Lp-norm corruptions to augment the training and test data of image classifiers. We adapt an approach from the field of adversarial robustness to assess the model robustness to imperceptible random corruptions. We empirically and theoretically investigate whether robustness is transferable across different Lp-norms and derive conclusions on which Lp-norm corruptions a model should be trained and evaluated on. We find that training data augmentation with L0-norm corruptions improves corruption robustness while maintaining accuracy compared to standard training and when applied on top of selected state-of-the-art data augmentation techniques.
[[2305.04989] Knowledge Graph Guided Semantic Evaluation of Language Models For User Trust](http://arxiv.org/abs/2305.04989) #robustA fundamental question in natural language processing is - what kind of language structure and semantics is the language model capturing? Graph formats such as knowledge graphs are easy to evaluate as they explicitly express language semantics and structure. This study evaluates the semantics encoded in the self-attention transformers by leveraging explicit knowledge graph structures. We propose novel metrics to measure the reconstruction error when providing graph path sequences from a knowledge graph and trying to reproduce/reconstruct the same from the outputs of the self-attention transformer models. The opacity of language models has an immense bearing on societal issues of trust and explainable decision outcomes. Our findings suggest that language models are models of stochastic control processes for plausible language pattern generation. However, they do not ascribe object and concept-level meaning and semantics to the learned stochastic patterns such as those described in knowledge graphs. Furthermore, to enable robust evaluation of concept understanding by language models, we construct and make public an augmented language understanding benchmark built on the General Language Understanding Evaluation (GLUE) benchmark. This has significant application-level user trust implications as stochastic patterns without a strong sense of meaning cannot be trusted in high-stakes applications.
[[2305.04990] Explanation-based Finetuning Makes Models More Robust to Spurious Cues](http://arxiv.org/abs/2305.04990) #robustLarge Language Models (LLMs) are so powerful that they sometimes learn correlations between labels and features that are irrelevant to the task, leading to poor generalization on out-of-distribution data. We propose explanation-based finetuning as a novel and general approach to mitigate LLMs' reliance on spurious correlations. Unlike standard finetuning where the model only predicts the answer given the input, we finetune the model to additionally generate a free-text explanation supporting its answer. To evaluate our method, we finetune the model on artificially constructed training sets containing different types of spurious cues, and test it on a test set without these cues. Compared to standard finetuning, our method makes models remarkably more robust against spurious cues in terms of accuracy drop across four classification tasks: ComVE (+1.2), CREAK (+9.1), e-SNLI (+15.4), and SBIC (+6.5). Moreover, our method works equally well with explanations generated by the model, implying its applicability to more datasets without human-written explanations.
[[2305.05079] A Unified Evaluation Framework for Novelty Detection and Accommodation in NLP with an Instantiation in Authorship Attribution](http://arxiv.org/abs/2305.05079) #robustState-of-the-art natural language processing models have been shown to achieve remarkable performance in 'closed-world' settings where all the labels in the evaluation set are known at training time. However, in real-world settings, 'novel' instances that do not belong to any known class are often observed. This renders the ability to deal with novelties crucial. To initiate a systematic research in this important area of 'dealing with novelties', we introduce 'NoveltyTask', a multi-stage task to evaluate a system's performance on pipelined novelty 'detection' and 'accommodation' tasks. We provide mathematical formulation of NoveltyTask and instantiate it with the authorship attribution task that pertains to identifying the correct author of a given text. We use Amazon reviews corpus and compile a large dataset (consisting of 250k instances across 200 authors/labels) for NoveltyTask. We conduct comprehensive experiments and explore several baseline methods for the task. Our results show that the methods achieve considerably low performance making the task challenging and leaving sufficient room for improvement. Finally, we believe our work will encourage research in this underexplored area of dealing with novelties, an important step en route to developing robust systems.
[[2305.05214] Utilizing Lexical Similarity to Enable Zero-Shot Machine Translation for Extremely Low-resource Languages](http://arxiv.org/abs/2305.05214) #robustWe address the task of machine translation from an extremely low-resource language (LRL) to English using cross-lingual transfer from a closely related high-resource language (HRL). For many of these languages, no parallel corpora are available, even monolingual corpora are limited and representations in pre-trained sequence-to-sequence models are absent. These factors limit the benefits of cross-lingual transfer from shared embedding spaces in multilingual models. However, many extremely LRLs have a high level of lexical similarity with related HRLs. We utilize this property by injecting character and character-span noise into the training data of the HRL prior to learning the vocabulary. This serves as a regularizer which makes the model more robust to lexical divergences between the HRL and LRL and better facilitates cross-lingual transfer. On closely related HRL and LRL pairs from multiple language families, we observe that our method significantly outperforms the baseline MT as well as approaches proposed previously to address cross-lingual transfer between closely related languages. We also show that the proposed character-span noise injection performs better than the unigram-character noise injection.
[[2305.05271] Robust Acoustic and Semantic Contextual Biasing in Neural Transducers for Speech Recognition](http://arxiv.org/abs/2305.05271) #robustAttention-based contextual biasing approaches have shown significant improvements in the recognition of generic and/or personal rare-words in End-to-End Automatic Speech Recognition (E2E ASR) systems like neural transducers. These approaches employ cross-attention to bias the model towards specific contextual entities injected as bias-phrases to the model. Prior approaches typically relied on subword encoders for encoding the bias phrases. However, subword tokenizations are coarse and fail to capture granular pronunciation information which is crucial for biasing based on acoustic similarity. In this work, we propose to use lightweight character representations to encode fine-grained pronunciation features to improve contextual biasing guided by acoustic similarity between the audio and the contextual entities (termed acoustic biasing). We further integrate pretrained neural language model (NLM) based encoders to encode the utterance's semantic context along with contextual entities to perform biasing informed by the utterance's semantic context (termed semantic biasing). Experiments using a Conformer Transducer model on the Librispeech dataset show a 4.62% - 9.26% relative WER improvement on different biasing list sizes over the baseline contextual model when incorporating our proposed acoustic and semantic biasing approach. On a large-scale in-house dataset, we observe 7.91% relative WER improvement compared to our baseline model. On tail utterances, the improvements are even more pronounced with 36.80% and 23.40% relative WER improvements on Librispeech rare words and an in-house testset respectively.
[[2305.05280] VCSUM: A Versatile Chinese Meeting Summarization Dataset](http://arxiv.org/abs/2305.05280) #robustCompared to news and chat summarization, the development of meeting summarization is hugely decelerated by the limited data. To this end, we introduce a versatile Chinese meeting summarization dataset, dubbed VCSum, consisting of 239 real-life meetings, with a total duration of over 230 hours. We claim our dataset is versatile because we provide the annotations of topic segmentation, headlines, segmentation summaries, overall meeting summaries, and salient sentences for each meeting transcript. As such, the dataset can adapt to various summarization tasks or methods, including segmentation-based summarization, multi-granularity summarization and retrieval-then-generate summarization. Our analysis confirms the effectiveness and robustness of VCSum. We also provide a set of benchmark models regarding different downstream summarization tasks on VCSum to facilitate further research. The dataset and code will be released at \url{https://github.com/hahahawu/VCSum}.
[[2305.05295] Boosting Zero-shot Cross-lingual Retrieval by Training on Artificially Code-Switched Data](http://arxiv.org/abs/2305.05295) #robustTransferring information retrieval (IR) models from a high-resource language (typically English) to other languages in a zero-shot fashion has become a widely adopted approach. In this work, we show that the effectiveness of zero-shot rankers diminishes when queries and documents are present in different languages. Motivated by this, we propose to train ranking models on artificially code-switched data instead, which we generate by utilizing bilingual lexicons. To this end, we experiment with lexicons induced from (1) cross-lingual word embeddings and (2) parallel Wikipedia page titles. We use the mMARCO dataset to extensively evaluate reranking models on 36 language pairs spanning Monolingual IR (MoIR), Cross-lingual IR (CLIR), and Multilingual IR (MLIR). Our results show that code-switching can yield consistent and substantial gains of 5.1 MRR@10 in CLIR and 3.9 MRR@10 in MLIR, while maintaining stable performance in MoIR. Encouragingly, the gains are especially pronounced for distant languages (up to 2x absolute gain). We further show that our approach is robust towards the ratio of code-switched tokens and also extends to unseen languages. Our results demonstrate that training on code-switched data is a cheap and effective way of generalizing zero-shot rankers for cross-lingual and multilingual retrieval.
[[2305.05116] Communication-Robust Multi-Agent Learning by Adaptable Auxiliary Multi-Agent Adversary Generation](http://arxiv.org/abs/2305.05116) #robustCommunication can promote coordination in cooperative Multi-Agent Reinforcement Learning (MARL). Nowadays, existing works mainly focus on improving the communication efficiency of agents, neglecting that real-world communication is much more challenging as there may exist noise or potential attackers. Thus the robustness of the communication-based policies becomes an emergent and severe issue that needs more exploration. In this paper, we posit that the ego system trained with auxiliary adversaries may handle this limitation and propose an adaptable method of Multi-Agent Auxiliary Adversaries Generation for robust Communication, dubbed MA3C, to obtain a robust communication-based policy. In specific, we introduce a novel message-attacking approach that models the learning of the auxiliary attacker as a cooperative problem under a shared goal to minimize the coordination ability of the ego system, with which every information channel may suffer from distinct message attacks. Furthermore, as naive adversarial training may impede the generalization ability of the ego system, we design an attacker population generation approach based on evolutionary learning. Finally, the ego system is paired with an attacker population and then alternatively trained against the continuously evolving attackers to improve its robustness, meaning that both the ego system and the attackers are adaptable. Extensive experiments on multiple benchmarks indicate that our proposed MA3C provides comparable or better robustness and generalization ability than other baselines.
[[2305.05230] FedNoRo: Towards Noise-Robust Federated Learning by Addressing Class Imbalance and Label Noise Heterogeneity](http://arxiv.org/abs/2305.05230) #robustFederated noisy label learning (FNLL) is emerging as a promising tool for privacy-preserving multi-source decentralized learning. Existing research, relying on the assumption of class-balanced global data, might be incapable to model complicated label noise, especially in medical scenarios. In this paper, we first formulate a new and more realistic federated label noise problem where global data is class-imbalanced and label noise is heterogeneous, and then propose a two-stage framework named FedNoRo for noise-robust federated learning. Specifically, in the first stage of FedNoRo, per-class loss indicators followed by Gaussian Mixture Model are deployed for noisy client identification. In the second stage, knowledge distillation and a distance-aware aggregation function are jointly adopted for noise-robust federated model updating. Experimental results on the widely-used ICH and ISIC2019 datasets demonstrate the superiority of FedNoRo against the state-of-the-art FNLL methods for addressing class imbalance and label noise heterogeneity in real-world FL scenarios.
[[2305.05392] On the Relation between Sharpness-Aware Minimization and Adversarial Robustness](http://arxiv.org/abs/2305.05392) #robustWe propose a novel understanding of Sharpness-Aware Minimization (SAM) in the context of adversarial robustness. In this paper, we point out that both SAM and adversarial training (AT) can be viewed as specific feature perturbations, which improve adversarial robustness. However, we note that SAM and AT are distinct in terms of perturbation strength, leading to different accuracy and robustness trade-offs. We provide theoretical evidence for these claims in a simplified model with rigorous mathematical proofs. Furthermore, we conduct experiment to demonstrate that only utilizing SAM can achieve superior adversarial robustness compared to standard training, which is an unexpected benefit. As adversarial training can suffer from a decrease in clean accuracy, we show that using SAM alone can improve robustness without sacrificing clean accuracy. Code is available at https://github.com/weizeming/SAM_AT.
[[2305.05448] Robust Implicit Regularization via Weight Normalization](http://arxiv.org/abs/2305.05448) #robustOverparameterized models may have many interpolating solutions; implicit regularization refers to the hidden preference of a particular optimization method towards a certain interpolating solution among the many. A by now established line of work has shown that (stochastic) gradient descent tends to have an implicit bias towards low rank and/or sparse solutions when used to train deep linear networks, explaining to some extent why overparameterized neural network models trained by gradient descent tend to have good generalization performance in practice. However, existing theory for square-loss objectives often requires very small initialization of the trainable weights, which is at odds with the larger scale at which weights are initialized in practice for faster convergence and better generalization performance. In this paper, we aim to close this gap by incorporating and analyzing gradient descent with weight normalization, where the weight vector is reparamterized in terms of polar coordinates, and gradient descent is applied to the polar coordinates. By analyzing key invariants of the gradient flow and using Lojasiewicz's Theorem, we show that weight normalization also has an implicit bias towards sparse solutions in the diagonal linear model, but that in contrast to plain gradient descent, weight normalization enables a robust bias that persists even if the weights are initialized at practically large scale. Experiments suggest that the gains in both convergence speed and robustness of the implicit bias are improved dramatically by using weight normalization in overparameterized diagonal linear network models.
[[2305.05161] Child Palm-ID: Contactless Palmprint Recognition for Children](http://arxiv.org/abs/2305.05161) #biometricEffective distribution of nutritional and healthcare aid for children, particularly infants and toddlers, in some of the least developed and most impoverished countries of the world, is a major problem due to the lack of reliable identification documents. Biometric authentication technology has been investigated to address child recognition in the absence of reliable ID documents. We present a mobile-based contactless palmprint recognition system, called Child Palm-ID, which meets the requirements of usability, hygiene, cost, and accuracy for child recognition. Using a contactless child palmprint database, Child-PalmDB1, consisting of 19,158 images from 1,020 unique palms (in the age range of 6 mos. to 48 mos.), we report a TAR=94.11% @ FAR=0.1%. The proposed Child Palm-ID system is also able to recognize adults, achieving a TAR=99.4% on the CASIA contactless palmprint database and a TAR=100% on the COEP contactless adult palmprint database, both @ FAR=0.1%. These accuracies are competitive with the SOTA provided by COTS systems. Despite these high accuracies, we show that the TAR for time-separated child-palmprints is only 78.1% @ FAR=0.1%.
[[2305.05293] On the Limitations of Model Stealing with Uncertainty Quantification Models](http://arxiv.org/abs/2305.05293) #stealModel stealing aims at inferring a victim model's functionality at a fraction of the original training cost. While the goal is clear, in practice the model's architecture, weight dimension, and original training data can not be determined exactly, leading to mutual uncertainty during stealing. In this work, we explicitly tackle this uncertainty by generating multiple possible networks and combining their predictions to improve the quality of the stolen model. For this, we compare five popular uncertainty quantification models in a model stealing task. Surprisingly, our results indicate that the considered models only lead to marginal improvements in terms of label agreement (i.e., fidelity) to the stolen model. To find the cause of this, we inspect the diversity of the model's prediction by looking at the prediction variance as a function of training iterations. We realize that during training, the models tend to have similar predictions, indicating that the network diversity we wanted to leverage using uncertainty quantification models is not (high) enough for improvements on the model stealing task.
[[2305.05200] LSAS: Lightweight Sub-attention Strategy for Alleviating Attention Bias Problem](http://arxiv.org/abs/2305.05200) #extractionIn computer vision, the performance of deep neural networks (DNNs) is highly related to the feature extraction ability, i.e., the ability to recognize and focus on key pixel regions in an image. However, in this paper, we quantitatively and statistically illustrate that DNNs have a serious attention bias problem on many samples from some popular datasets: (1) Position bias: DNNs fully focus on label-independent regions; (2) Range bias: The focused regions from DNN are not completely contained in the ideal region. Moreover, we find that the existing self-attention modules can alleviate these biases to a certain extent, but the biases are still non-negligible. To further mitigate them, we propose a lightweight sub-attention strategy (LSAS), which utilizes high-order sub-attention modules to improve the original self-attention modules. The effectiveness of LSAS is demonstrated by extensive experiments on widely-used benchmark datasets and popular attention networks. We release our code to help other researchers to reproduce the results of LSAS~\footnote{https://github.com/Qrange-group/LSAS}.
[[2305.05296] Mediapipe and CNNs for Real-Time ASL Gesture Recognition](http://arxiv.org/abs/2305.05296) #extractionThis research paper describes a realtime system for identifying American Sign Language (ASL) movements that employs modern computer vision and machine learning approaches. The suggested method makes use of the Mediapipe library for feature extraction and a Convolutional Neural Network (CNN) for ASL gesture classification. The testing results show that the suggested system can detect all ASL alphabets with an accuracy of 99.95%, indicating its potential for use in communication devices for people with hearing impairments. The proposed approach can also be applied to additional sign languages with similar hand motions, potentially increasing the quality of life for people with hearing loss. Overall, the study demonstrates the effectiveness of using Mediapipe and CNN for real-time sign language recognition, making a significant contribution to the field of computer vision and machine learning.
[[2305.05003] Revisiting Relation Extraction in the era of Large Language Models](http://arxiv.org/abs/2305.05003) #extractionRelation extraction (RE) is the core NLP task of inferring semantic relationships between entities from text. Standard supervised RE techniques entail training modules to tag tokens comprising entity spans and then predict the relationship between them. Recent work has instead treated the problem as a \emph{sequence-to-sequence} task, linearizing relations between entities as target strings to be generated conditioned on the input. Here we push the limits of this approach, using larger language models (GPT-3 and Flan-T5 large) than considered in prior work and evaluating their performance on standard RE tasks under varying levels of supervision. We address issues inherent to evaluating generative approaches to RE by doing human evaluations, in lieu of relying on exact matching. Under this refined evaluation, we find that: (1) Few-shot prompting with GPT-3 achieves near SOTA performance, i.e., roughly equivalent to existing fully supervised models; (2) Flan-T5 is not as capable in the few-shot setting, but supervising and fine-tuning it with Chain-of-Thought (CoT) style explanations (generated via GPT-3) yields SOTA results. We release this model as a new baseline for RE tasks.
[[2305.05119] Flexible Job Shop Scheduling via Dual Attention Network Based Reinforcement Learning](http://arxiv.org/abs/2305.05119) #extractionFlexible manufacturing has given rise to complex scheduling problems such as the flexible job shop scheduling problem (FJSP). In FJSP, operations can be processed on multiple machines, leading to intricate relationships between operations and machines. Recent works have employed deep reinforcement learning (DRL) to learn priority dispatching rules (PDRs) for solving FJSP. However, the quality of solutions still has room for improvement relative to that by the exact methods such as OR-Tools. To address this issue, this paper presents a novel end-to-end learning framework that weds the merits of self-attention models for deep feature extraction and DRL for scalable decision-making. The complex relationships between operations and machines are represented precisely and concisely, for which a dual-attention network (DAN) comprising several interconnected operation message attention blocks and machine message attention blocks is proposed. The DAN exploits the complicated relationships to construct production-adaptive operation and machine features to support high-quality decisionmaking. Experimental results using synthetic data as well as public benchmarks corroborate that the proposed approach outperforms both traditional PDRs and the state-of-the-art DRL method. Moreover, it achieves results comparable to exact methods in certain cases and demonstrates favorable generalization ability to large-scale and real-world unseen FJSP tasks.
[[2305.05644] Towards Building the Federated GPT: Federated Instruction Tuning](http://arxiv.org/abs/2305.05644) #federateWhile ``instruction-tuned" generative large language models (LLMs) have demonstrated an impressive ability to generalize to new tasks, the training phases heavily rely on large amounts of diverse and high-quality instruction data (such as ChatGPT and GPT-4). Unfortunately, acquiring high-quality data, especially when it comes to human-written data, can pose significant challenges both in terms of cost and accessibility. Moreover, concerns related to privacy can further limit access to such data, making the process of obtaining it a complex and nuanced undertaking. Consequently, this hinders the generality of the tuned models and may restrict their effectiveness in certain contexts. To tackle this issue, our study introduces a new approach called Federated Instruction Tuning (FedIT), which leverages federated learning (FL) as the learning framework for the instruction tuning of LLMs. This marks the first exploration of FL-based instruction tuning for LLMs. This is especially important since text data is predominantly generated by end users. Therefore, it is imperative to design and adapt FL approaches to effectively leverage these users' diverse instructions stored on local devices, while preserving privacy and ensuring data security. In the current paper, by conducting widely used GPT-4 auto-evaluation, we demonstrate that by exploiting the heterogeneous and diverse sets of instructions on the client's end with the proposed framework FedIT, we improved the performance of LLMs compared to centralized training with only limited local instructions. Further, in this paper, we developed a Github repository named Shepherd. This repository offers a foundational framework for exploring federated fine-tuning of LLMs using heterogeneous instructions across diverse categories.
[[2305.04979] FedHB: Hierarchical Bayesian Federated Learning](http://arxiv.org/abs/2305.04979) #federateWe propose a novel hierarchical Bayesian approach to Federated Learning (FL), where our model reasonably describes the generative process of clients' local data via hierarchical Bayesian modeling: constituting random variables of local models for clients that are governed by a higher-level global variate. Interestingly, the variational inference in our Bayesian model leads to an optimisation problem whose block-coordinate descent solution becomes a distributed algorithm that is separable over clients and allows them not to reveal their own private data at all, thus fully compatible with FL. We also highlight that our block-coordinate algorithm has particular forms that subsume the well-known FL algorithms including Fed-Avg and Fed-Prox as special cases. Beyond introducing novel modeling and derivations, we also offer convergence analysis showing that our block-coordinate FL algorithm converges to an (local) optimum of the objective at the rate of $O(1/\sqrt{t})$, the same rate as regular (centralised) SGD, as well as the generalisation error analysis where we prove that the test error of our model on unseen data is guaranteed to vanish as we increase the training data size, thus asymptotically optimal.
[[2305.05090] Performative Federated Learning: A Solution to Model-Dependent and Heterogeneous Distribution Shifts](http://arxiv.org/abs/2305.05090) #federateWe consider a federated learning (FL) system consisting of multiple clients and a server, where the clients aim to collaboratively learn a common decision model from their distributed data. Unlike the conventional FL framework that assumes the client's data is static, we consider scenarios where the clients' data distributions may be reshaped by the deployed decision model. In this work, we leverage the idea of distribution shift mappings in performative prediction to formalize this model-dependent data distribution shift and propose a performative federated learning framework. We first introduce necessary and sufficient conditions for the existence of a unique performative stable solution and characterize its distance to the performative optimal solution. Then we propose the performative FedAvg algorithm and show that it converges to the performative stable solution at a rate of O(1/T) under both full and partial participation schemes. In particular, we use novel proof techniques and show how the clients' heterogeneity influences the convergence. Numerical results validate our analysis and provide valuable insights into real-world applications.
[[2305.05110] Semi-Supervised Federated Learning for Keyword Spotting](http://arxiv.org/abs/2305.05110) #federateKeyword Spotting (KWS) is a critical aspect of audio-based applications on mobile devices and virtual assistants. Recent developments in Federated Learning (FL) have significantly expanded the ability to train machine learning models by utilizing the computational and private data resources of numerous distributed devices. However, existing FL methods typically require that devices possess accurate ground-truth labels, which can be both expensive and impractical when dealing with local audio data. In this study, we first demonstrate the effectiveness of Semi-Supervised Federated Learning (SSL) and FL for KWS. We then extend our investigation to Semi-Supervised Federated Learning (SSFL) for KWS, where devices possess completely unlabeled data, while the server has access to a small amount of labeled data. We perform numerical analyses using state-of-the-art SSL, FL, and SSFL techniques to demonstrate that the performance of KWS models can be significantly improved by leveraging the abundant unlabeled heterogeneous data available on devices.
[[2305.05118] Federated Learning Operations Made Simple with Flame](http://arxiv.org/abs/2305.05118) #federateDistributed machine learning approaches, including a broad class of federated learning techniques, present a number of benefits when deploying machine learning applications over widely distributed infrastructures. To realize the expected benefits, however, introduces substantial operational challenges due to required application and configuration-level changes related to deployment-specific details. Such complexities can be greatly reduced by introducing higher-level abstractions -- role and channel -- using which federated learning applications are described as Topology Abstraction Graphs (TAGs). TAGs decouple the ML application logic from the underlying deployment details, making it possible to specialize the application deployment, thus reducing development effort and paving the way for improved automation and tuning. We present Flame, the first system that supports these abstractions, and demonstrate its benefits for several use cases.
[[2305.05221] BARA: Efficient Incentive Mechanism with Online Reward Budget Allocation in Cross-Silo Federated Learning](http://arxiv.org/abs/2305.05221) #federateFederated learning (FL) is a prospective distributed machine learning framework that can preserve data privacy. In particular, cross-silo FL can complete model training by making isolated data islands of different organizations collaborate with a parameter server (PS) via exchanging model parameters for multiple communication rounds. In cross-silo FL, an incentive mechanism is indispensable for motivating data owners to contribute their models to FL training. However, how to allocate the reward budget among different rounds is an essential but complicated problem largely overlooked by existing works. The challenge of this problem lies in the opaque feedback between reward budget allocation and model utility improvement of FL, making the optimal reward budget allocation complicated. To address this problem, we design an online reward budget allocation algorithm using Bayesian optimization named BARA (\underline{B}udget \underline{A}llocation for \underline{R}everse \underline{A}uction). Specifically, BARA can model the complicated relationship between reward budget allocation and final model accuracy in FL based on historical training records so that the reward budget allocated to each communication round is dynamically optimized so as to maximize the final model utility. We further incorporate the BARA algorithm into reverse auction-based incentive mechanisms to illustrate its effectiveness. Extensive experiments are conducted on real datasets to demonstrate that BARA significantly outperforms competitive baselines by improving model utility with the same amount of reward budget.
[[2305.05257] Survey of Federated Learning Models for Spatial-Temporal Mobility Applications](http://arxiv.org/abs/2305.05257) #federateFederated learning involves training statistical models over edge devices such as mobile phones such that the training data is kept local. Federated Learning (FL) can serve as an ideal candidate for training spatial temporal models that rely on heterogeneous and potentially massive numbers of participants while preserving the privacy of highly sensitive location data. However, there are unique challenges involved with transitioning existing spatial temporal models to decentralized learning. In this survey paper, we review the existing literature that has proposed FL-based models for predicting human mobility, traffic prediction, community detection, location-based recommendation systems, and other spatial-temporal tasks. We describe the metrics and datasets these works have been using and create a baseline of these approaches in comparison to the centralized settings. Finally, we discuss the challenges of applying spatial-temporal models in a decentralized setting and by highlighting the gaps in the literature we provide a road map and opportunities for the research community.
[[2305.05314] CAMIL: Context-Aware Multiple Instance Learning for Whole Slide Image Classification](http://arxiv.org/abs/2305.05314) #interpretabilityCancer diagnoses typically involve human pathologists examining whole slide images (WSIs) of tissue section biopsies to identify tumor cells and their subtypes. However, artificial intelligence (AI)-based models, particularly weakly supervised approaches, have recently emerged as viable alternatives. Weakly supervised approaches often use image subsections or tiles as input, with the overall classification of the WSI based on attention scores assigned to each tile. However, this method overlooks the potential for false positives/negatives because tumors can be heterogeneous, with cancer and normal cells growing in patterns larger than a single tile. Such errors at the tile level could lead to misclassification at the tumor level. To address this limitation, we developed a novel deep learning pooling operator called CHARM (Contrastive Histopathology Attention Resolved Models). CHARM leverages the dependencies among single tiles within a WSI and imposes contextual constraints as prior knowledge to multiple instance learning models. We tested CHARM on the subtyping of non-small cell lung cancer (NSLC) and lymph node (LN) metastasis, and the results demonstrated its superiority over other state-of-the-art weakly supervised classification algorithms. Furthermore, CHARM facilitates interpretability by visualizing regions of attention.
[[2305.05349] Towards the Characterization of Representations Learned via Capsule-based Network Architectures](http://arxiv.org/abs/2305.05349) #interpretabilityCapsule Networks (CapsNets) have been re-introduced as a more compact and interpretable alternative to standard deep neural networks. While recent efforts have proved their compression capabilities, to date, their interpretability properties have not been fully assessed. Here, we conduct a systematic and principled study towards assessing the interpretability of these types of networks. Moreover, we pay special attention towards analyzing the level to which part-whole relationships are indeed encoded within the learned representation. Our analysis in the MNIST, SVHN, PASCAL-part and CelebA datasets suggest that the representations encoded in CapsNets might not be as disentangled nor strictly related to parts-whole relationships as is commonly stated in the literature.
[[2305.05505] Recursions Are All You Need: Towards Efficient Deep Unfolding Networks](http://arxiv.org/abs/2305.05505) #interpretabilityThe use of deep unfolding networks in compressive sensing (CS) has seen wide success as they provide both simplicity and interpretability. However, since most deep unfolding networks are iterative, this incurs significant redundancies in the network. In this work, we propose a novel recursion-based framework to enhance the efficiency of deep unfolding models. First, recursions are used to effectively eliminate the redundancies in deep unfolding networks. Secondly, we randomize the number of recursions during training to decrease the overall training time. Finally, to effectively utilize the power of recursions, we introduce a learnable unit to modulate the features of the model based on both the total number of iterations and the current iteration index. To evaluate the proposed framework, we apply it to both ISTA-Net+ and COAST. Extensive testing shows that our proposed framework allows the network to cut down as much as 75% of its learnable parameters while mostly maintaining its performance, and at the same time, it cuts around 21% and 42% from the training time for ISTA-Net+ and COAST respectively. Moreover, when presented with a limited training dataset, the recursive models match or even outperform their respective non-recursive baseline. Codes and pretrained models are available at https://github.com/Rawwad-Alhejaili/Recursions-Are-All-You-Need .
[[2305.05111] When a CBR in Hand is Better than Twins in the Bush](http://arxiv.org/abs/2305.05111) #interpretabilityAI methods referred to as interpretable are often discredited as inaccurate by supporters of the existence of a trade-off between interpretability and accuracy. In many problem contexts however this trade-off does not hold. This paper discusses a regression problem context to predict flight take-off delays where the most accurate data regression model was trained via the XGBoost implementation of gradient boosted decision trees. While building an XGB-CBR Twin and converting the XGBoost feature importance into global weights in the CBR model, the resultant CBR model alone provides the most accurate local prediction, maintains the global importance to provide a global explanation of the model, and offers the most interpretable representation for local explanations. This resultant CBR model becomes a benchmark of accuracy and interpretability for this problem context, and hence it is used to evaluate the two additive feature attribute methods SHAP and LIME to explain the XGBoost regression model. The results with respect to local accuracy and feature attribution lead to potentially valuable future work.
[[2305.05562] SkelEx and BoundEx: Natural Visualization of ReLU Neural Networks](http://arxiv.org/abs/2305.05562) #interpretabilityDespite their limited interpretability, weights and biases are still the most popular encoding of the functions learned by ReLU Neural Networks (ReLU NNs). That is why we introduce SkelEx, an algorithm to extract a skeleton of the membership functions learned by ReLU NNs, making those functions easier to interpret and analyze. To the best of our knowledge, this is the first work that considers linear regions from the perspective of critical points. As a natural follow-up, we also introduce BoundEx, which is the first analytical method known to us to extract the decision boundary from the realization of a ReLU NN. Both of those methods introduce very natural visualization tool for ReLU NNs trained on low-dimensional data.
[[2305.05077] Atmospheric Turbulence Correction via Variational Deep Diffusion](http://arxiv.org/abs/2305.05077) #diffusionAtmospheric Turbulence (AT) correction is a challenging restoration task as it consists of two distortions: geometric distortion and spatially variant blur. Diffusion models have shown impressive accomplishments in photo-realistic image synthesis and beyond. In this paper, we propose a novel deep conditional diffusion model under a variational inference framework to solve the AT correction problem. We use this framework to improve performance by learning latent prior information from the input and degradation processes. We use the learned information to further condition the diffusion model. Experiments are conducted in a comprehensive synthetic AT dataset. We show that the proposed framework achieves good quantitative and qualitative results.
[[2305.05189] SUR-adapter: Enhancing Text-to-Image Pre-trained Diffusion Models with Large Language Models](http://arxiv.org/abs/2305.05189) #diffusionDiffusion models, which have emerged to become popular text-to-image generation models, can produce high-quality and content-rich images guided by textual prompts. However, there are limitations to semantic understanding and commonsense reasoning in existing models when the input prompts are concise narrative, resulting in low-quality image generation. To improve the capacities for narrative prompts, we propose a simple-yet-effective parameter-efficient fine-tuning approach called the Semantic Understanding and Reasoning adapter (SUR-adapter) for pre-trained diffusion models. To reach this goal, we first collect and annotate a new dataset SURD which consists of more than 57,000 semantically corrected multi-modal samples. Each sample contains a simple narrative prompt, a complex keyword-based prompt, and a high-quality image. Then, we align the semantic representation of narrative prompts to the complex prompts and transfer knowledge of large language models (LLMs) to our SUR-adapter via knowledge distillation so that it can acquire the powerful semantic understanding and reasoning capabilities to build a high-quality textual semantic representation for text-to-image generation. We conduct experiments by integrating multiple LLMs and popular pre-trained diffusion models to show the effectiveness of our approach in enabling diffusion models to understand and reason concise natural language without image quality degradation. Our approach can make text-to-image diffusion models easier to use with better user experience, which demonstrates our approach has the potential for further advancing the development of user-friendly text-to-image generation models by bridging the semantic gap between simple narrative prompts and complex keyword-based prompts.
[[2305.05464] Style-A-Video: Agile Diffusion for Arbitrary Text-based Video Style Transfer](http://arxiv.org/abs/2305.05464) #diffusionLarge-scale text-to-video diffusion models have demonstrated an exceptional ability to synthesize diverse videos. However, due to the lack of extensive text-to-video datasets and the necessary computational resources for training, directly applying these models for video stylization remains difficult. Also, given that the noise addition process on the input content is random and destructive, fulfilling the style transfer task's content preservation criteria is challenging. This paper proposes a zero-shot video stylization method named Style-A-Video, which utilizes a generative pre-trained transformer with an image latent diffusion model to achieve a concise text-controlled video stylization. We improve the guidance condition in the denoising process, establishing a balance between artistic expression and structure preservation. Furthermore, to decrease inter-frame flicker and avoid the formation of additional artifacts, we employ a sampling optimization and a temporal consistency module. Extensive experiments show that we can attain superior content preservation and stylistic performance while incurring less consumption than previous solutions. Code will be available at https://github.com/haha-lisa/Style-A-Video.
[[2305.04961] Joint Moment Retrieval and Highlight Detection Via Natural Language Queries](http://arxiv.org/abs/2305.04961) #transformerVideo summarization has become an increasingly important task in the field of computer vision due to the vast amount of video content available on the internet. In this project, we propose a new method for natural language query based joint video summarization and highlight detection using multi-modal transformers. This approach will use both visual and audio cues to match a user's natural language query to retrieve the most relevant and interesting moments from a video. Our approach employs multiple recent techniques used in Vision Transformers (ViTs) to create a transformer-like encoder-decoder model. We evaluated our approach on multiple datasets such as YouTube Highlights and TVSum to demonstrate the flexibility of our proposed method.
[[2305.05177] Hybrid Transformer and CNN Attention Network for Stereo Image Super-resolution](http://arxiv.org/abs/2305.05177) #transformerMulti-stage strategies are frequently employed in image restoration tasks. While transformer-based methods have exhibited high efficiency in single-image super-resolution tasks, they have not yet shown significant advantages over CNN-based methods in stereo super-resolution tasks. This can be attributed to two key factors: first, current single-image super-resolution transformers are unable to leverage the complementary stereo information during the process; second, the performance of transformers is typically reliant on sufficient data, which is absent in common stereo-image super-resolution algorithms. To address these issues, we propose a Hybrid Transformer and CNN Attention Network (HTCAN), which utilizes a transformer-based network for single-image enhancement and a CNN-based network for stereo information fusion. Furthermore, we employ a multi-patch training strategy and larger window sizes to activate more input pixels for super-resolution. We also revisit other advanced techniques, such as data augmentation, data ensemble, and model ensemble to reduce overfitting and data bias. Finally, our approach achieved a score of 23.90dB and emerged as the winner in Track 1 of the NTIRE 2023 Stereo Image Super-Resolution Challenge.
[[2305.05534] Integrating Holistic and Local Information to Estimate Emotional Reaction Intensity](http://arxiv.org/abs/2305.05534) #transformerVideo-based Emotional Reaction Intensity (ERI) estimation measures the intensity of subjects' reactions to stimuli along several emotional dimensions from videos of the subject as they view the stimuli. We propose a multi-modal architecture for video-based ERI combining video and audio information. Video input is encoded spatially first, frame-by-frame, combining features encoding holistic aspects of the subjects' facial expressions and features encoding spatially localized aspects of their expressions. Input is then combined across time: from frame-to-frame using gated recurrent units (GRUs), then globally by a transformer. We handle variable video length with a regression token that accumulates information from all frames into a fixed-dimensional vector independent of video length. Audio information is handled similarly: spectral information extracted within each frame is integrated across time by a cascade of GRUs and a transformer with regression token. The video and audio regression tokens' outputs are merged by concatenation, then input to a final fully connected layer producing intensity estimates. Our architecture achieved excellent performance on the Hume-Reaction dataset in the ERI Esimation Challenge of the Fifth Competition on Affective Behavior Analysis in-the-Wild (ABAW5). The Pearson Correlation Coefficients between estimated and subject self-reported scores, averaged across all emotions, were 0.455 on the validation dataset and 0.4547 on the test dataset, well above the baselines. The transformer's self-attention mechanism enables our architecture to focus on the most critical video frames regardless of length. Ablation experiments establish the advantages of combining holistic/local features and of multi-modal integration. Code available at https://github.com/HKUST-NISL/ABAW5.
[[2305.05546] ColonMapper: topological mapping and localization for colonoscopy](http://arxiv.org/abs/2305.05546) #transformerMapping and localization in endoluminal cavities from colonoscopies or gastroscopies has to overcome the challenge of significant shape and illumination changes between reobservations of the same endoluminal location. Instead of geometrical maps that strongly rely on a fixed scene geometry, topological maps are more adequate because they focus on visual place recognition, i.e. the capability to determine if two video shots are imaging the same location. We propose a topological mapping and localization system able to operate on real human colonoscopies. The map is a graph where each node codes a colon location by a set of real images of that location. The edges represent traversability between two nodes. For close-in-time images, where scene changes are minor, place recognition can be successfully managed with the recent transformers-based image-matching algorithms. However, under long-term changes -- such as different colonoscopies of the same patient -- feature-based matching fails. To address this, we propose a GeM global descriptor able to achieve high recall with significant changes in the scene. The addition of a Bayesian filter processing the map graph boosts the accuracy of the long-term place recognition, enabling relocalization in a previously built map. In the experiments, we construct a map during the withdrawal phase of a first colonoscopy. Subsequently, we prove the ability to relocalize within this map during a second colonoscopy of the same patient two weeks later. Code and models will be available upon acceptance.
[[2305.05583] Group Activity Recognition via Dynamic Composition and Interaction](http://arxiv.org/abs/2305.05583) #transformerPrevious group activity recognition approaches were limited to reasoning using human relations or finding important subgroups and tended to ignore indispensable group composition and human-object interactions. This absence makes a partial interpretation of the scene and increases the interference of irrelevant actions on the results. Therefore, we propose our DynamicFormer with Dynamic composition Module (DcM) and Dynamic interaction Module (DiM) to model relations and locations of persons and discriminate the contribution of participants, respectively. Our findings on group composition and human-object interaction inspire our core idea. Group composition tells us the location of people and their relations inside the group, while interaction reflects the relation between humans and objects outside the group. We utilize spatial and temporal encoders in DcM to model our dynamic composition and build DiM to explore interaction with a novel GCN, which has a transformer inside to consider the temporal neighbors of human/object. Also, a Multi-level Dynamic Integration is employed to integrate features from different levels. We conduct extensive experiments on two public datasets and show that our method achieves state-of-the-art.
[[2305.05651] SwinIA: Self-Supervised Blind-Spot Image Denoising with Zero Convolutions](http://arxiv.org/abs/2305.05651) #transformerThe essence of self-supervised image denoising is to restore the signal from the noisy image alone. State-of-the-art solutions for this task rely on the idea of masking pixels and training a fully-convolutional neural network to impute them. This most often requires multiple forward passes, information about the noise model, and intricate regularization functions. In this paper, we propose a Swin Transformer-based Image Autoencoder (SwinIA), the first convolution-free architecture for self-supervised denoising. It can be trained end-to-end with a simple mean squared error loss without masking and does not require any prior knowledge about clean data or noise distribution. Despite its simplicity, SwinIA establishes state-of-the-art on several common benchmarks.
[[2305.04928] A transformer-based method for zero and few-shot biomedical named entity recognition](http://arxiv.org/abs/2305.04928) #transformerSupervised named entity recognition (NER) in the biomedical domain is dependent on large sets of annotated texts with the given named entities, whose creation can be time-consuming and expensive. Furthermore, the extraction of new entities often requires conducting additional annotation tasks and retraining the model. To address these challenges, this paper proposes a transformer-based method for zero- and few-shot NER in the biomedical domain. The method is based on transforming the task of multi-class token classification into binary token classification (token contains the searched entity or does not contain the searched entity) and pre-training on a larger amount of datasets and biomedical entities, from where the method can learn semantic relations between the given and potential classes. We have achieved average F1 scores of 35.44% for zero-shot NER, 50.10% for one-shot NER, 69.94% for 10-shot NER, and 79.51% for 100-shot NER on 9 diverse evaluated biomedical entities with PubMedBERT fine-tuned model. The results demonstrate the effectiveness of the proposed method for recognizing new entities with limited examples, with comparable or better results from the state-of-the-art zero- and few-shot NER methods.
[[2305.05061] Coherent Wave Dynamics and Language Generation of a Generative Pre-trained Transformer](http://arxiv.org/abs/2305.05061) #transformerLarge Language Models (LLMs), such as the Generative Pretrained Transformer (GPT), have achieved tremendous success in various language tasks, but their emergent abilities have also raised many questions, concerns, and challenges that need to be addressed. To gain a better understanding of the models' inner mechanisms, we analyze the hidden state and channel wave dynamics in a small GPT, focusing on the coherence of wave patterns in terms of cross-channel correlation and individual auto-correlation. Our findings suggest that wave dynamics offer consistent and repeatable intrinsic oscillation modes, along with context-aware plasticity and expressiveness in language generation. By analyzing wave patterns, coherence, and clustering, we provide a systematic way to identify and interpret the functionality of the hidden state channels, paving the way to understand and control higher-level language pattern formation. In addition, we investigate the Poisson statistics of spelling errors in text sequence generation across various levels of model training and observe a phase-transition-like process. As coherence builds up, there is a competition between the generation of correct and misspelled words. However, once the model is adequately trained and significant coherence has emerged, the coherent process becomes strong enough to effectively suppress spelling errors, preventing the cascade amplification of defects. The distribution of correct spellings transitions from Poissonian to Sub-Poissonian, while the distribution of misspellings shows the opposite trend. By leveraging concepts and techniques from quantum physics, we gain novel insights into the dynamics of the small GPT. This approach can be extended to larger language models that exhibit more complex coherent language patterns, opening up opportunities to interpret their emergent capabilities and develop more specialized models.
[[2305.05325] Detection of depression on social networks using transformers and ensembles](http://arxiv.org/abs/2305.05325) #transformerAs the impact of technology on our lives is increasing, we witness increased use of social media that became an essential tool not only for communication but also for sharing information with community about our thoughts and feelings. This can be observed also for people with mental health disorders such as depression where they use social media for expressing their thoughts and asking for help. This opens a possibility to automatically process social media posts and detect signs of depression. We build several large pre-trained language model based classifiers for depression detection from social media posts. Besides fine-tuning BERT, RoBERTA, BERTweet, and mentalBERT were also construct two types of ensembles. We analyze the performance of our models on two data sets of posts from social platforms Reddit and Twitter, and investigate also the performance of transfer learning across the two data sets. The results show that transformer ensembles improve over the single transformer-based classifiers.
[[2305.05480] Investigating the effect of sub-word segmentation on the performance of transformer language models](http://arxiv.org/abs/2305.05480) #transformerWe would like to explore how morphemes can affect the performance of a language model. We trained GPT-2 and Bert model with StateMorph for both Finnish and Russian, which is a morpheme segmenting algorithm. As a comparison, we also trained a model with BPE and Morfessor. Our preliminary result shows that StateMorph can help the model to converge more efficiently and achieve a better validation score.
[[2305.05465] The emergence of clusters in self-attention dynamics](http://arxiv.org/abs/2305.05465) #transformerViewing Transformers as interacting particle systems, we describe the geometry of learned representations when the weights are not time dependent. We show that particles, representing tokens, tend to cluster toward particular limiting objects as time tends to infinity. The type of limiting object that emerges depends on the spectrum of the value matrix. Additionally, in the one-dimensional case we prove that the self-attention matrix converges to a low-rank Boolean matrix. The combination of these results mathematically confirms the empirical observation made by Vaswani et al. \cite{vaswani2017attention} that \emph{leaders} appear in a sequence of tokens when processed by Transformers.
[[2305.05351] GPT-NAS: Neural Architecture Search with the Generative Pre-Trained Model](http://arxiv.org/abs/2305.05351) #generativeNeural Architecture Search (NAS) has emerged as one of the effective methods to design the optimal neural network architecture automatically. Although neural architectures have achieved human-level performances in several tasks, few of them are obtained from the NAS method. The main reason is the huge search space of neural architectures, making NAS algorithms inefficient. This work presents a novel architecture search algorithm, called GPT-NAS, that optimizes neural architectures by Generative Pre-Trained (GPT) model. In GPT-NAS, we assume that a generative model pre-trained on a large-scale corpus could learn the fundamental law of building neural architectures. Therefore, GPT-NAS leverages the generative pre-trained (GPT) model to propose reasonable architecture components given the basic one. Such an approach can largely reduce the search space by introducing prior knowledge in the search process. Extensive experimental results show that our GPT-NAS method significantly outperforms seven manually designed neural architectures and thirteen architectures provided by competing NAS methods. In addition, our ablation study indicates that the proposed algorithm improves the performance of finely tuned neural architectures by up to about 12% compared to those without GPT, further demonstrating its effectiveness in searching neural architectures.
[[2305.05580] Fashion CUT: Unsupervised domain adaptation for visual pattern classification in clothes using synthetic data and pseudo-labels](http://arxiv.org/abs/2305.05580) #generativeAccurate product information is critical for e-commerce stores to allow customers to browse, filter, and search for products. Product data quality is affected by missing or incorrect information resulting in poor customer experience. While machine learning can be used to correct inaccurate or missing information, achieving high performance on fashion image classification tasks requires large amounts of annotated data, but it is expensive to generate due to labeling costs. One solution can be to generate synthetic data which requires no manual labeling. However, training a model with a dataset of solely synthetic images can lead to poor generalization when performing inference on real-world data because of the domain shift. We introduce a new unsupervised domain adaptation technique that converts images from the synthetic domain into the real-world domain. Our approach combines a generative neural network and a classifier that are jointly trained to produce realistic images while preserving the synthetic label information. We found that using real-world pseudo-labels during training helps the classifier to generalize in the real-world domain, reducing the synthetic bias. We successfully train a visual pattern classification model in the fashion domain without real-world annotations. Experiments show that our method outperforms other unsupervised domain adaptation algorithms.
[[2305.05402] Consistent Text Categorization using Data Augmentation in e-Commerce](http://arxiv.org/abs/2305.05402) #generativeThe categorization of massive e-Commerce data is a crucial, well-studied task, which is prevalent in industrial settings. In this work, we aim to improve an existing product categorization model that is already in use by a major web company, serving multiple applications. At its core, the product categorization model is a text classification model that takes a product title as an input and outputs the most suitable category out of thousands of available candidates. Upon a closer inspection, we found inconsistencies in the labeling of similar items. For example, minor modifications of the product title pertaining to colors or measurements majorly impacted the model's output. This phenomenon can negatively affect downstream recommendation or search applications, leading to a sub-optimal user experience.
To address this issue, we propose a new framework for consistent text categorization. Our goal is to improve the model's consistency while maintaining its production-level performance. We use a semi-supervised approach for data augmentation and presents two different methods for utilizing unlabeled samples. One method relies directly on existing catalogs, while the other uses a generative model. We compare the pros and cons of each approach and present our experimental results.
[[2305.05001] GersteinLab at MEDIQA-Chat 2023: Clinical Note Summarization from Doctor-Patient Conversations through Fine-tuning and In-context Learning](http://arxiv.org/abs/2305.05001) #large language modelThis paper presents our contribution to the MEDIQA-2023 Dialogue2Note shared task, encompassing both subtask A and subtask B. We approach the task as a dialogue summarization problem and implement two distinct pipelines: (a) a fine-tuning of a pre-trained dialogue summarization model and GPT-3, and (b) few-shot in-context learning (ICL) using a large language model, GPT-4. Both methods achieve excellent results in terms of ROUGE-1 F1, BERTScore F1 (deberta-xlarge-mnli), and BLEURT, with scores of 0.4011, 0.7058, and 0.5421, respectively. Additionally, we predict the associated section headers using RoBERTa and SciBERT based classification models. Our team ranked fourth among all teams, while each team is allowed to submit three runs as part of their submission. We also utilize expert annotations to demonstrate that the notes generated through the ICL GPT-4 are better than all other baselines. The code for our submission is available.
[[2305.05027] Web Content Filtering through knowledge distillation of Large Language Models](http://arxiv.org/abs/2305.05027) #large language modelWe introduce a state-of-the-art approach for URL categorization that leverages the power of Large Language Models (LLMs) to address the primary objectives of web content filtering: safeguarding organizations from legal and ethical risks, limiting access to high-risk or suspicious websites, and fostering a secure and professional work environment. Our method utilizes LLMs to generate accurate classifications and then employs established knowledge distillation techniques to create smaller, more specialized student models tailored for web content filtering. Distillation results in a student model with a 9\% accuracy rate improvement in classifying websites, sourced from customer telemetry data collected by a large security vendor, into 30 distinct content categories based on their URLs, surpassing the current state-of-the-art approach. Our student model matches the performance of the teacher LLM with 175 times less parameters, allowing the model to be used for in-line scanning of large volumes of URLs, and requires 3 orders of magnitude less manually labeled training data than the current state-of-the-art approach. Depending on the specific use case, the output generated by our approach can either be directly returned or employed as a pre-filter for more resource-intensive operations involving website images or HTML.
[[2305.05050] ANALOGICAL -- A New Benchmark for Analogy of Long Text for Large Language Models](http://arxiv.org/abs/2305.05050) #large language modelOver the past decade, analogies, in the form of word-level analogies, have played a significant role as an intrinsic measure of evaluating the quality of word embedding methods such as word2vec. Modern large language models (LLMs), however, are primarily evaluated on extrinsic measures based on benchmarks such as GLUE and SuperGLUE, and there are only a few investigations on whether LLMs can draw analogies between long texts. In this paper, we present ANALOGICAL, a new benchmark to intrinsically evaluate LLMs across a taxonomy of analogies of long text with six levels of complexity -- (i) word, (ii) word vs. sentence, (iii) syntactic, (iv) negation, (v) entailment, and (vi) metaphor. Using thirteen datasets and three different distance measures, we evaluate the abilities of eight LLMs in identifying analogical pairs in the semantic vector space (e.g., "I can speak two languages" should be closer to "I am bilingual" while "I like chocolate" and "I do not like chocolate" should be orthogonal). Our evaluation finds that it is increasingly challenging for LLMs to identify analogies when going up the analogy taxonomy.
[[2305.05054] Dreams Are More "Predictable'' Than You Think](http://arxiv.org/abs/2305.05054) #large language modelA consistent body of evidence suggests that dream reports significantly vary from other types of textual transcripts with respect to semantic content. Furthermore, it appears to be a widespread belief in the dream/sleep research community that dream reports constitute rather ``unique'' strings of text. This might be a notable issue for the growing amount of approaches using natural language processing (NLP) tools to automatically analyse dream reports, as they largely rely on neural models trained on non-dream corpora scraped from the web. In this work, I will adopt state-of-the-art (SotA) large language models (LLMs), to study if and how dream reports deviate from other human-generated text strings, such as Wikipedia. Results show that, taken as a whole, DreamBank does not deviate from Wikipedia. Moreover, on average, single dream reports are significantly more predictable than Wikipedia articles. Preliminary evidence suggests that word count, gender, and visual impairment can significantly shape how predictable a dream report can appear to the model.
[[2305.05176] FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance](http://arxiv.org/abs/2305.05176) #large language modelThere is a rapidly growing number of large language models (LLMs) that users can query for a fee. We review the cost associated with querying popular LLM APIs, e.g. GPT-4, ChatGPT, J1-Jumbo, and find that these models have heterogeneous pricing structures, with fees that can differ by two orders of magnitude. In particular, using LLMs on large collections of queries and text can be expensive. Motivated by this, we outline and discuss three types of strategies that users can exploit to reduce the inference cost associated with using LLMs: 1) prompt adaptation, 2) LLM approximation, and 3) LLM cascade. As an example, we propose FrugalGPT, a simple yet flexible instantiation of LLM cascade which learns which combinations of LLMs to use for different queries in order to reduce cost and improve accuracy. Our experiments show that FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost. The ideas and findings presented here lay a foundation for using LLMs sustainably and efficiently.
[[2305.05181] MoT: Pre-thinking and Recalling Enable ChatGPT to Self-Improve with Memory-of-Thoughts](http://arxiv.org/abs/2305.05181) #large language modelLarge Language Models have shown impressive abilities on various tasks. However, fundamentally improving them depends on high-quality datasets or computationally expensive fine-tuning. On the contrary, human can easily improve themselves by thinking and memory, without external resources. In this paper, we propose a framework, MoT, to let the LLM self-improve through Memory of Thoughts, without annotated datasets and parameter updates. Specifically, the framework is divided into two stages: 1. before the test stage, we let the LLM pre-think on the unlabeled dataset and save the high-confidence thoughts as external memory; 2. during inference, given a test question, we let the LLM recall relevant memory to help itself reason and answer it. Experimental results show that the proposed framework can help ChatGPT significantly improve its abilities in math reasoning, commonsense reasoning, factual reasoning and natural language inference. Further analyses show that each component contributes critically to the improvements.
[[2305.05252] Distilling Script Knowledge from Large Language Models for Constrained Language Planning](http://arxiv.org/abs/2305.05252) #large language modelIn everyday life, humans often plan their actions by following step-by-step instructions in the form of goal-oriented scripts. Previous work has exploited language models (LMs) to plan for abstract goals of stereotypical activities (e.g., "make a cake"), but leaves more specific goals with multi-facet constraints understudied (e.g., "make a cake for diabetics"). In this paper, we define the task of constrained language planning for the first time. We propose an overgenerate-then-filter approach to improve large language models (LLMs) on this task, and use it to distill a novel constrained language planning dataset, CoScript, which consists of 55,000 scripts. Empirical results demonstrate that our method significantly improves the constrained language planning ability of LLMs, especially on constraint faithfulness. Furthermore, CoScript is demonstrated to be quite effective in endowing smaller LMs with constrained language planning ability.
[[2305.05364] Large Language Model Programs](http://arxiv.org/abs/2305.05364) #large language modelIn recent years, large pre-trained language models (LLMs) have demonstrated the ability to follow instructions and perform novel tasks from a few examples. The possibility to parameterise an LLM through such in-context examples widens their capability at a much lower cost than finetuning. We extend this line of reasoning and present a method which further expands the capabilities of an LLM by embedding it within an algorithm or program. To demonstrate the benefits of this approach, we present an illustrative example of evidence-supported question-answering. We obtain a 6.4\% improvement over the chain of thought baseline through a more algorithmic approach without any finetuning. Furthermore, we highlight recent work from this perspective and discuss the advantages and disadvantages in comparison to the standard approaches.
[[2305.05410] Large Language Models Need Holistically Thought in Medical Conversational QA](http://arxiv.org/abs/2305.05410) #large language modelThe medical conversational question answering (CQA) system aims at providing a series of professional medical services to improve the efficiency of medical care. Despite the success of large language models (LLMs) in complex reasoning tasks in various fields, such as mathematics, logic, and commonsense QA, they still need to improve with the increased complexity and specialization of the medical field. This is because medical CQA tasks require not only strong medical reasoning, but also the ability to think broadly and deeply. In this paper, to address these challenges in medical CQA tasks that need to be considered and understood in many aspects, we propose the Holistically Thought (HoT) method, which is designed to guide the LLMs to perform the diffused and focused thinking for generating high-quality medical responses. The proposed HoT method has been evaluated through automated and manual assessments in three different medical CQA datasets containing the English and Chinese languages. The extensive experimental results show that our method can produce more correctness, professional, and considerate answers than several state-of-the-art (SOTA) methods, manifesting its effectiveness. Our code in https://github.com/WENGSYX/HoT.
[[2305.05609] The Case Records of ChatGPT: Language Models and Complex Clinical Questions](http://arxiv.org/abs/2305.05609) #large language modelBackground: Artificial intelligence language models have shown promise in various applications, including assisting with clinical decision-making as demonstrated by strong performance of large language models on medical licensure exams. However, their ability to solve complex, open-ended cases, which may be representative of clinical practice, remains unexplored. Methods: In this study, the accuracy of large language AI models GPT4 and GPT3.5 in diagnosing complex clinical cases was investigated using published Case Records of the Massachusetts General Hospital. A total of 50 cases requiring a diagnosis and diagnostic test published from January 1, 2022 to April 16, 2022 were identified. For each case, models were given a prompt requesting the top three specific diagnoses and associated diagnostic tests, followed by case text, labs, and figure legends. Model outputs were assessed in comparison to the final clinical diagnosis and whether the model-predicted test would result in a correct diagnosis. Results: GPT4 and GPT3.5 accurately provided the correct diagnosis in 26% and 22% of cases in one attempt, and 46% and 42% within three attempts, respectively. GPT4 and GPT3.5 provided a correct essential diagnostic test in 28% and 24% of cases in one attempt, and 44% and 50% within three attempts, respectively. No significant differences were found between the two models, and multiple trials with identical prompts using the GPT3.5 model provided similar results. Conclusions: In summary, these models demonstrate potential usefulness in generating differential diagnoses but remain limited in their ability to provide a single unifying diagnosis in complex, open-ended cases. Future research should focus on evaluating model performance in larger datasets of open-ended clinical challenges and exploring potential human-AI collaboration strategies to enhance clinical decision-making.
[[2305.05132] Dual flow fusion model for concrete surface crack segmentation](http://arxiv.org/abs/2305.05132) #segmentationCracks and other diseases are important factors that threaten the safe operation of transportation infrastructure. Traditional manual detection and ultrasonic instrument detection consume a lot of time and resource costs. With the development of deep learning technology, many deep learning models are widely used in actual visual segmentation tasks. The detection method based on the deep learning model has the advantages of high detection accuracy, fast detection speed and simple operation. However, the crack segmentation based on deep learning has problems such as sensitivity to background noise, rough edges, and lack of robustness. Therefore, this paper proposes a fissure segmentation model based on two-stream fusion, which simultaneously inputs images into two designed processing streams to independently extract long-distance dependent and local detail features, and realizes adaptive prediction through a dual-head mechanism. At the same time, a new interactive fusion mechanism is proposed to guide the complementarity of different levels of features to realize the location and identification of cracks in complex backgrounds. Finally, we propose an edge optimization method to improve segmentation accuracy. Experiments have proved that the F1 value of the segmentation results on the DeepCrack[1] public dataset reached 93.7%, and the IOU value reached 86.6%; the F1 value of the segmentation results on the CRACK500[2] dataset reached 78.1%, and the IOU value reached 66.0%.
[[2305.05154] Multi-Granularity Denoising and Bidirectional Alignment for Weakly Supervised Semantic Segmentation](http://arxiv.org/abs/2305.05154) #segmentationWeakly supervised semantic segmentation (WSSS) models relying on class activation maps (CAMs) have achieved desirable performance comparing to the non-CAMs-based counterparts. However, to guarantee WSSS task feasible, we need to generate pseudo labels by expanding the seeds from CAMs which is complex and time-consuming, thus hindering the design of efficient end-to-end (single-stage) WSSS approaches. To tackle the above dilemma, we resort to the off-the-shelf and readily accessible saliency maps for directly obtaining pseudo labels given the image-level class labels. Nevertheless, the salient regions may contain noisy labels and cannot seamlessly fit the target objects, and saliency maps can only be approximated as pseudo labels for simple images containing single-class objects. As such, the achieved segmentation model with these simple images cannot generalize well to the complex images containing multi-class objects. To this end, we propose an end-to-end multi-granularity denoising and bidirectional alignment (MDBA) model, to alleviate the noisy label and multi-class generalization issues. Specifically, we propose the online noise filtering and progressive noise detection modules to tackle image-level and pixel-level noise, respectively. Moreover, a bidirectional alignment mechanism is proposed to reduce the data distribution gap at both input and output space with simple-to-complex image synthesis and complex-to-simple adversarial learning. MDBA can reach the mIoU of 69.5\% and 70.2\% on validation and test sets for the PASCAL VOC 2012 dataset. The source codes and models have been made available at \url{https://github.com/NUST-Machine-Intelligence-Laboratory/MDBA}.
[[2305.05421] DC3DCD: unsupervised learning for multiclass 3D point cloud change detection](http://arxiv.org/abs/2305.05421) #segmentationIn a constant evolving world, change detection is of prime importance to keep updated maps. To better sense areas with complex geometry (urban areas in particular), considering 3D data appears to be an interesting alternative to classical 2D images. In this context, 3D point clouds (PCs) obtained by LiDAR or photogrammetry are very interesting. While recent studies showed the considerable benefit of using deep learning-based methods to detect and characterize changes into raw 3D PCs, these studies rely on large annotated training data to obtain accurate results. The collection of these annotations are tricky and time-consuming. The availability of unsupervised or weakly supervised approaches is then of prime interest. In this paper, we propose an unsupervised method, called DeepCluster 3D Change Detection (DC3DCD), to detect and categorize multiclass changes at point level. We classify our approach in the unsupervised family given the fact that we extract in a completely unsupervised way a number of clusters associated with potential changes. Let us precise that in the end of the process, the user has only to assign a label to each of these clusters to derive the final change map. Our method builds upon the DeepCluster approach, originally designed for image classification, to handle complex raw 3D PCs and perform change segmentation task. An assessment of the method on both simulated and real public dataset is provided. The proposed method allows to outperform fully-supervised traditional machine learning algorithm and to be competitive with fully-supervised deep learning networks applied on rasterization of 3D PCs with a mean of IoU over classes of change of 57.06% and 66.69% for the simulated and the real datasets, respectively.
[[2305.05490] Real-time instance segmentation with polygons using an Intersection-over-Union loss](http://arxiv.org/abs/2305.05490) #segmentationPredicting a binary mask for an object is more accurate but also more computationally expensive than a bounding box. Polygonal masks as developed in CenterPoly can be a good compromise. In this paper, we improve over CenterPoly by enhancing the classical regression L1 loss with a novel region-based loss and a novel order loss, as well as with a new training process for the vertices prediction head. Moreover, the previous methods that predict polygonal masks use different coordinate systems, but it is not clear if one is better than another, if we abstract the architecture requirement. We therefore investigate their impact on the prediction. We also use a new evaluation protocol with oracle predictions for the detection head, to further isolate the segmentation process and better compare the polygonal masks with binary masks. Our instance segmentation method is trained and tested with challenging datasets containing urban scenes, with a high density of road users. Experiments show, in particular, that using a combination of a regression loss and a region-based loss allows significant improvements on the Cityscapes and IDD test set compared to CenterPoly. Moreover the inference stage remains fast enough to reach real-time performance with an average of 0.045 s per frame for 2048$\times$1024 images on a single RTX 2070 GPU. The code is available $\href{https://github.com/KatiaJDL/CenterPoly-v2}{\text{here}}$.
[[2305.05511] Self-supervised dense representation learning for live-cell microscopy with time arrow prediction](http://arxiv.org/abs/2305.05511) #segmentationState-of-the-art object detection and segmentation methods for microscopy images rely on supervised machine learning, which requires laborious manual annotation of training data. Here we present a self-supervised method based on time arrow prediction pre-training that learns dense image representations from raw, unlabeled live-cell microscopy videos. Our method builds upon the task of predicting the correct order of time-flipped image regions via a single-image feature extractor and a subsequent time arrow prediction head. We show that the resulting dense representations capture inherently time-asymmetric biological processes such as cell divisions on a pixel-level. We furthermore demonstrate the utility of these representations on several live-cell microscopy datasets for detection and segmentation of dividing cells, as well as for cell state classification. Our method outperforms supervised methods, particularly when only limited ground truth annotations are available as is commonly the case in practice. We provide code at https://github.com/weigertlab/tarrow.
[[2305.05610] Can point cloud networks learn statistical shape models of anatomies?](http://arxiv.org/abs/2305.05610) #segmentationStatistical Shape Modeling (SSM) is a valuable tool for investigating and quantifying anatomical variations within populations of anatomies. However, traditional correspondence-based SSM generation methods require a time-consuming re-optimization process each time a new subject is added to the cohort, making the inference process prohibitive for clinical research. Additionally, they require complete geometric proxies (e.g., high-resolution binary volumes or surface meshes) as input shapes to construct the SSM. Unordered 3D point cloud representations of shapes are more easily acquired from various medical imaging practices (e.g., thresholded images and surface scanning). Point cloud deep networks have recently achieved remarkable success in learning permutation-invariant features for different point cloud tasks (e.g., completion, semantic segmentation, classification). However, their application to learning SSM from point clouds is to-date unexplored. In this work, we demonstrate that existing point cloud encoder-decoder-based completion networks can provide an untapped potential for SSM, capturing population-level statistical representations of shapes while reducing the inference burden and relaxing the input requirement. We discuss the limitations of these techniques to the SSM application and suggest future improvements. Our work paves the way for further exploration of point cloud deep learning for SSM, a promising avenue for advancing shape analysis literature and broadening SSM to diverse use cases.