PyTorchKR Papers

최신 논문

총 4,526편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2302.07455

A lightweight network for photovoltaic cell defect detection in electroluminescence images based on neural architecture search and knowledge distillation

Nowadays, the rapid development of photovoltaic(PV) power stations requires increasingly reliable maintenance and fault diagnosis of PV modules in the field. Due to the effectiveness, convolutional neural network (CNN) has been widely used in the existing automatic defect detection of PV cells. However, the parameters of these CNN-based models are very large, which require stringent hardware resources and it is difficult to be applied in actual industrial projects. To solve these problems, we propose a novel lightweight high-performance model for automatic defect detection of PV cells in electroluminescence(EL) images based on neural architecture search and knowledge distillation. To auto-design an effective lightweight model, we introduce neural architecture search to the field of PV cell defect classification for the first time. Since the defect can be any size, we design a proper search structure of network to better exploit the multi-scale characteristic. To improve the overall performance of the searched lightweight model, we further transfer the knowledge learned by the existing pre-trained large-scale model based on knowledge distillation. Different kinds of knowledge are exploited and transferred, including attention information, feature information, logit information and task-oriented information. Experiments have demonstrated that the proposed model achieves the state-of-the-art performance on the public PV cell dataset of EL images under online data augmentation with accuracy of 91.74% and the parameters of 1.85M. The proposed lightweight high-performance model can be easily deployed to the end devices of the actual industrial projects and retain the accuracy.

arXiv:2302.07215공식 코드 · Papers with Code

Multi-teacher knowledge distillation as an effective method for compressing ensembles of neural networks

Deep learning has contributed greatly to many successes in artificial intelligence in recent years. Today, it is possible to train models that have thousands of layers and hundreds of billions of parameters. Large-scale deep models have achieved great success, but the enormous computational complexity and gigantic storage requirements make it extremely difficult to implement them in real-time applications. On the other hand, the size of the dataset is still a real problem in many domains. Data are often missing, too expensive, or impossible to obtain for other reasons. Ensemble learning is partially a solution to the problem of small datasets and overfitting. However, ensemble learning in its basic version is associated with a linear increase in computational complexity. We analyzed the impact of the ensemble decision-fusion mechanism and checked various methods of sharing the decisions including voting algorithms. We used the modified knowledge distillation framework as a decision-fusion mechanism which allows in addition compressing of the entire ensemble model into a weight space of a single model. We showed that knowledge distillation can aggregate knowledge from multiple teachers in only one student model and, with the same computational complexity, obtain a better-performing model compared to a model trained in the standard manner. We have developed our own method for mimicking the responses of all teachers at the same time, simultaneously. We tested these solutions on several benchmark datasets. In the end, we presented a wide application use of the efficient multi-teacher knowledge distillation framework. In the first example, we used knowledge distillation to develop models that could automate corrosion detection on aircraft fuselage. The second example describes detection of smoke on observation cameras in order to counteract wildfires in forests.

arXiv:2302.06898

Take a Prior from Other Tasks for Severe Blur Removal

Recovering clear structures from severely blurry inputs is a challenging problem due to the large movements between the camera and the scene. Although some works apply segmentation maps on human face images for deblurring, they cannot handle natural scenes because objects and degradation are more complex, and inaccurate segmentation maps lead to a loss of details. For general scene deblurring, the feature space of the blurry image and corresponding sharp image under the high-level vision task is closer, which inspires us to rely on other tasks (e.g. classification) to learn a comprehensive prior in severe blur removal cases. We propose a cross-level feature learning strategy based on knowledge distillation to learn the priors, which include global contexts and sharp local structures for recovering potential details. In addition, we propose a semantic prior embedding layer with multi-level aggregation and semantic attention transformation to integrate the priors effectively. We introduce the proposed priors to various models, including the UNet and other mainstream deblurring baselines, leading to better performance on severe blur removal. Extensive experiments on natural image deblurring benchmarks and real-world images, such as GoPro and RealBlur datasets, demonstrate our method's effectiveness and generalization ability.

arXiv:2302.06637공식 코드 · Papers with Code

PerAda: Parameter-Efficient Federated Learning Personalization with Generalization Guarantees

Personalized Federated Learning (pFL) has emerged as a promising solution to tackle data heterogeneity across clients in FL. However, existing pFL methods either (1) introduce high communication and computation costs or (2) overfit to local data, which can be limited in scope, and are vulnerable to evolved test samples with natural shifts. In this paper, we propose PerAda, a parameter-efficient pFL framework that reduces communication and computational costs and exhibits superior generalization performance, especially under test-time distribution shifts. PerAda reduces the costs by leveraging the power of pretrained models and only updates and communicates a small number of additional parameters from adapters. PerAda has good generalization since it regularizes each client's personalized adapter with a global adapter, while the global adapter uses knowledge distillation to aggregate generalized information from all clients. Theoretically, we provide generalization bounds to explain why PerAda improves generalization, and we prove its convergence to stationary points under non-convex settings. Empirically, PerAda demonstrates competitive personalized performance (+4.85% on CheXpert) and enables better out-of-distribution generalization (+5.23% on CIFAR-10-C) on different datasets across natural and medical domains compared with baselines, while only updating 12.6% of parameters per model based on the adapter. Our code is available at https://github.com/NVlabs/PerAda.

arXiv:2302.06078

NYCU-TWO at Memotion 3: Good Foundation, Good Teacher, then you have Good Meme Analysis

This paper presents a robust solution to the Memotion 3.0 Shared Task. The goal of this task is to classify the emotion and the corresponding intensity expressed by memes, which are usually in the form of images with short captions on social media. Understanding the multi-modal features of the given memes will be the key to solving the task. In this work, we use CLIP to extract aligned image-text features and propose a novel meme sentiment analysis framework, consisting of a Cooperative Teaching Model (CTM) for Task A and a Cascaded Emotion Classifier (CEC) for Tasks B&C. CTM is based on the idea of knowledge distillation, and can better predict the sentiment of a given meme in Task A; CEC can leverage the emotion intensity suggestion from the prediction of Task C to classify the emotion more precisely in Task B. Experiments show that we achieved the 2nd place ranking for both Task A and Task B and the 4th place ranking for Task C, with weighted F1-scores of 0.342, 0.784, and 0.535 respectively. The results show the robustness and effectiveness of our framework. Our code is released at github.

arXiv:2302.05929

SCLIFD:Supervised Contrastive Knowledge Distillation for Incremental Fault Diagnosis under Limited Fault Data

Intelligent fault diagnosis has made extraordinary advancements currently. Nonetheless, few works tackle class-incremental learning for fault diagnosis under limited fault data, i.e., imbalanced and long-tailed fault diagnosis, which brings about various notable challenges. Initially, it is difficult to extract discriminative features from limited fault data. Moreover, a well-trained model must be retrained from scratch to classify the samples from new classes, thus causing a high computational burden and time consumption. Furthermore, the model may suffer from catastrophic forgetting when trained incrementally. Finally, the model decision is biased toward the new classes due to the class imbalance. The problems can consequently lead to performance degradation of fault diagnosis models. Accordingly, we introduce a supervised contrastive knowledge distillation for incremental fault diagnosis under limited fault data (SCLIFD) framework to address these issues, which extends the classical incremental classifier and representation learning (iCaRL) framework from three perspectives. Primarily, we adopt supervised contrastive knowledge distillation (KD) to enhance its representation learning capability under limited fault data. Moreover, we propose a novel prioritized exemplar selection method adaptive herding (AdaHerding) to restrict the increase of the computational burden, which is also combined with KD to alleviate catastrophic forgetting. Additionally, we adopt the cosine classifier to mitigate the adverse impact of class imbalance. We conduct extensive experiments on simulated and real-world industrial processes under different imbalance ratios. Experimental results show that our SCLIFD outperforms the existing methods by a large margin.

arXiv:2302.05637공식 코드 · Papers with Code

Dual Relation Knowledge Distillation for Object Detection

Knowledge distillation is an effective method for model compression. However, it is still a challenging topic to apply knowledge distillation to detection tasks. There are two key points resulting in poor distillation performance for detection tasks. One is the serious imbalance between foreground and background features, another one is that small object lacks enough feature representation. To solve the above issues, we propose a new distillation method named dual relation knowledge distillation (DRKD), including pixel-wise relation distillation and instance-wise relation distillation. The pixel-wise relation distillation embeds pixel-wise features in the graph space and applies graph convolution to capture the global pixel relation. By distilling the global pixel relation, the student detector can learn the relation between foreground and background features, and avoid the difficulty of distilling features directly for the feature imbalance issue. Besides, we find that instance-wise relation supplements valuable knowledge beyond independent features for small objects. Thus, the instance-wise relation distillation is designed, which calculates the similarity of different instances to obtain a relation matrix. More importantly, a relation filter module is designed to highlight valuable instance relations. The proposed dual relation knowledge distillation is general and can be easily applied for both one-stage and two-stage detectors. Our method achieves state-of-the-art performance, which improves Faster R-CNN based on ResNet50 from 38.4% to 41.6% mAP and improves RetinaNet based on ResNet50 from 37.4% to 40.3% mAP on COCO 2017.

arXiv:2302.06337공식 코드 · Papers with Code

Learning from Noisy Crowd Labels with Logics

This paper explores the integration of symbolic logic knowledge into deep neural networks for learning from noisy crowd labels. We introduce Logic-guided Learning from Noisy Crowd Labels (Logic-LNCL), an EM-alike iterative logic knowledge distillation framework that learns from both noisy labeled data and logic rules of interest. Unlike traditional EM methods, our framework contains a ``pseudo-E-step'' that distills from the logic rules a new type of learning target, which is then used in the ``pseudo-M-step'' for training the classifier. Extensive evaluations on two real-world datasets for text sentiment classification and named entity recognition demonstrate that the proposed framework improves the state-of-the-art and provides a new solution to learning from noisy crowd labels.

arXiv:2302.06195공식 코드 · Papers with Code

Exploring Navigation Maps for Learning-Based Motion Prediction

The prediction of surrounding agents' motion is a key for safe autonomous driving. In this paper, we explore navigation maps as an alternative to the predominant High Definition (HD) maps for learning-based motion prediction. Navigation maps provide topological and geometrical information on road-level, HD maps additionally have centimeter-accurate lane-level information. As a result, HD maps are costly and time-consuming to obtain, while navigation maps with near-global coverage are freely available. We describe an approach to integrate navigation maps into learning-based motion prediction models. To exploit locally available HD maps during training, we additionally propose a model-agnostic method for knowledge distillation. In experiments on the publicly available Argoverse dataset with navigation maps obtained from OpenStreetMap, our approach shows a significant improvement over not using a map at all. Combined with our method for knowledge distillation, we achieve results that are close to the original HD map-reliant models. Our publicly available navigation map API for Argoverse enables researchers to develop and evaluate their own approaches using navigation maps.

arXiv:2302.05324

SOCRATES: Text-based Human Search and Approach using a Robot Dog

In this paper, we propose a SOCratic model for Robots Approaching humans based on TExt System (SOCRATES) focusing on the human search and approach based on free-form textual description; the robot first searches for the target user, then the robot proceeds to approach in a human-friendly manner. In particular, textual descriptions are composed of appearance (e.g., wearing white shirts with black hair) and location clues (e.g., is a student who works with robots). We initially present a Human Search Socratic Model that connects large pre-trained models in the language domain to solve the downstream task, which is searching for the target person based on textual descriptions. Then, we propose a hybrid learning-based framework for generating target-cordial robotic motion to approach a person, consisting of a learning-from-demonstration module and a knowledge distillation module. We validate the proposed searching module via simulation using a virtual mobile robot as well as through real-world experiments involving participants and the Boston Dynamics Spot robot. Furthermore, we analyze the properties of the proposed approaching framework with human participants based on the Robotic Social Attributes Scale (RoSAS)

arXiv:2302.05213공식 코드 · Papers with Code

CEN-HDR: Computationally Efficient neural Network for real-time High Dynamic Range imaging

High dynamic range (HDR) imaging is still a challenging task in modern digital photography. Recent research proposes solutions that provide high-quality acquisition but at the cost of a very large number of operations and a slow inference time that prevent the implementation of these solutions on lightweight real-time systems. In this paper, we propose CEN-HDR, a new computationally efficient neural network by providing a novel architecture based on a light attention mechanism and sub-pixel convolution operations for real-time HDR imaging. We also provide an efficient training scheme by applying network compression using knowledge distillation. We performed extensive qualitative and quantitative comparisons to show that our approach produces competitive results in image quality while being faster than state-of-the-art solutions, allowing it to be practically deployed under real-time constraints. Experimental results show our method obtains a score of 43.04 mu-PSNR on the Kalantari2017 dataset with a framerate of 33 FPS using a Macbook M1 NPU.

arXiv:2302.04527공식 코드 · Papers with Code

Toward Extremely Lightweight Distracted Driver Recognition With Distillation-Based Neural Architecture Search and Knowledge Transfer

The number of traffic accidents has been continuously increasing in recent years worldwide. Many accidents are caused by distracted drivers, who take their attention away from driving. Motivated by the success of Convolutional Neural Networks (CNNs) in computer vision, many researchers developed CNN-based algorithms to recognize distracted driving from a dashcam and warn the driver against unsafe behaviors. However, current models have too many parameters, which is unfeasible for vehicle-mounted computing. This work proposes a novel knowledge-distillation-based framework to solve this problem. The proposed framework first constructs a high-performance teacher network by progressively strengthening the robustness to illumination changes from shallow to deep layers of a CNN. Then, the teacher network is used to guide the architecture searching process of a student network through knowledge distillation. After that, we use the teacher network again to transfer knowledge to the student network by knowledge distillation. Experimental results on the Statefarm Distracted Driver Detection Dataset and AUC Distracted Driver Dataset show that the proposed approach is highly effective for recognizing distracted driving behaviors from photos: (1) the teacher network's accuracy surpasses the previous best accuracy; (2) the student network achieves very high accuracy with only 0.42M parameters (around 55% of the previous most lightweight model). Furthermore, the student network architecture can be extended to a spatial-temporal 3D CNN for recognizing distracted driving from video clips. The 3D student network largely surpasses the previous best accuracy with only 2.03M parameters on the Drive&Act Dataset. The source code is available at https://github.com/Dichao-Liu/Lightweight_Distracted_Driver_Recognition_with_Distillation-Based_NAS_and_Knowledge_Transfer.

arXiv:2302.04308

Enhancing Modality-Agnostic Representations via Meta-Learning for Brain Tumor Segmentation

In medical vision, different imaging modalities provide complementary information. However, in practice, not all modalities may be available during inference or even training. Previous approaches, e.g., knowledge distillation or image synthesis, often assume the availability of full modalities for all patients during training; this is unrealistic and impractical due to the variability in data collection across sites. We propose a novel approach to learn enhanced modality-agnostic representations by employing a meta-learning strategy in training, even when only limited full modality samples are available. Meta-learning enhances partial modality representations to full modality representations by meta-training on partial modality data and meta-testing on limited full modality samples. Additionally, we co-supervise this feature enrichment by introducing an auxiliary adversarial learning branch. More specifically, a missing modality detector is used as a discriminator to mimic the full modality setting. Our segmentation framework significantly outperforms state-of-the-art brain tumor segmentation techniques in missing modality scenarios.

arXiv:2302.04725공식 코드 · Papers with Code

Lightweight Transformers for Clinical Natural Language Processing

Specialised pre-trained language models are becoming more frequent in NLP since they can potentially outperform models trained on generic texts. BioBERT and BioClinicalBERT are two examples of such models that have shown promise in medical NLP tasks. Many of these models are overparametrised and resource-intensive, but thanks to techniques like Knowledge Distillation (KD), it is possible to create smaller versions that perform almost as well as their larger counterparts. In this work, we specifically focus on development of compact language models for processing clinical texts (i.e. progress notes, discharge summaries etc). We developed a number of efficient lightweight clinical transformers using knowledge distillation and continual learning, with the number of parameters ranging from 15 million to 65 million. These models performed comparably to larger models such as BioBERT and ClinicalBioBERT and significantly outperformed other compact models trained on general or biomedical data. Our extensive evaluation was done across several standard datasets and covered a wide range of clinical text-mining tasks, including Natural Language Inference, Relation Extraction, Named Entity Recognition, and Sequence Classification. To our knowledge, this is the first comprehensive study specifically focused on creating efficient and compact transformers for clinical NLP tasks. The models and code used in this study can be found on our Huggingface profile at https://huggingface.co/nlpie and Github page at https://github.com/nlpie-research/Lightweight-Clinical-Transformers, respectively, promoting reproducibility of our results.

arXiv:2302.04112

An Empirical Study of Uniform-Architecture Knowledge Distillation in Document Ranking

Although BERT-based ranking models have been commonly used in commercial search engines, they are usually time-consuming for online ranking tasks. Knowledge distillation, which aims at learning a smaller model with comparable performance to a larger model, is a common strategy for reducing the online inference latency. In this paper, we investigate the effect of different loss functions for uniform-architecture distillation of BERT-based ranking models. Here "uniform-architecture" denotes that both teacher and student models are in cross-encoder architecture, while the student models include small-scaled pre-trained language models. Our experimental results reveal that the optimal distillation configuration for ranking tasks is much different than general natural language processing tasks. Specifically, when the student models are in cross-encoder architecture, a pairwise loss of hard labels is critical for training student models, whereas the distillation objectives of intermediate Transformer layers may hurt performance. These findings emphasize the necessity of carefully designing a distillation strategy (for cross-encoder student models) tailored for document ranking with pairwise training samples.

arXiv:2302.03806

SLaM: Student-Label Mixing for Distillation with Unlabeled Examples

Knowledge distillation with unlabeled examples is a powerful training paradigm for generating compact and lightweight student models in applications where the amount of labeled data is limited but one has access to a large pool of unlabeled data. In this setting, a large teacher model generates ``soft'' pseudo-labels for the unlabeled dataset which are then used for training the student model. Despite its success in a wide variety of applications, a shortcoming of this approach is that the teacher's pseudo-labels are often noisy, leading to impaired student performance. In this paper, we present a principled method for knowledge distillation with unlabeled examples that we call Student-Label Mixing (SLaM) and we show that it consistently improves over prior approaches by evaluating it on several standard benchmarks. Finally, we show that SLaM comes with theoretical guarantees; along the way we give an algorithm improving the best-known sample complexity for learning halfspaces with margin under random classification noise, and provide the first convergence analysis for so-called ``forward loss-adjustment" methods.

arXiv:2302.02069

Heterogeneous Federated Knowledge Graph Embedding Learning and Unlearning

Federated Learning (FL) recently emerges as a paradigm to train a global machine learning model across distributed clients without sharing raw data. Knowledge Graph (KG) embedding represents KGs in a continuous vector space, serving as the backbone of many knowledge-driven applications. As a promising combination, federated KG embedding can fully take advantage of knowledge learned from different clients while preserving the privacy of local data. However, realistic problems such as data heterogeneity and knowledge forgetting still remain to be concerned. In this paper, we propose FedLU, a novel FL framework for heterogeneous KG embedding learning and unlearning. To cope with the drift between local optimization and global convergence caused by data heterogeneity, we propose mutual knowledge distillation to transfer local knowledge to global, and absorb global knowledge back. Moreover, we present an unlearning method based on cognitive neuroscience, which combines retroactive interference and passive decay to erase specific knowledge from local clients and propagate to the global model by reusing knowledge distillation. We construct new datasets for assessing realistic performance of the state-of-the-arts. Extensive experiments show that FedLU achieves superior results in both link prediction and knowledge forgetting.

arXiv:2302.02845

Audio Representation Learning by Distilling Video as Privileged Information

Deep audio representation learning using multi-modal audio-visual data often leads to a better performance compared to uni-modal approaches. However, in real-world scenarios both modalities are not always available at the time of inference, leading to performance degradation by models trained for multi-modal inference. In this work, we propose a novel approach for deep audio representation learning using audio-visual data when the video modality is absent at inference. For this purpose, we adopt teacher-student knowledge distillation under the framework of learning using privileged information (LUPI). While the previous methods proposed for LUPI use soft-labels generated by the teacher, in our proposed method we use embeddings learned by the teacher to train the student network. We integrate our method in two different settings: sequential data where the features are divided into multiple segments throughout time, and non-sequential data where the entire features are treated as one whole segment. In the non-sequential setting both the teacher and student networks are comprised of an encoder component and a task header. We use the embeddings produced by the encoder component of the teacher to train the encoder of the student, while the task header of the student is trained using ground-truth labels. In the sequential setting, the networks have an additional aggregation component that is placed between the encoder and task header. We use two sets of embeddings produced by the encoder and aggregation component of the teacher to train the student. Similar to the non-sequential setting, the task header of the student network is trained using ground-truth labels. We test our framework on two different audio-visual tasks, namely speaker recognition and speech emotion recognition and show considerable improvements over sole audio-based recognition as well as prior works that use LUPI.

arXiv:2302.02108

Knowledge Distillation in Vision Transformers: A Critical Review

In Natural Language Processing (NLP), Transformers have already revolutionized the field by utilizing an attention-based encoder-decoder model. Recently, some pioneering works have employed Transformer-like architectures in Computer Vision (CV) and they have reported outstanding performance of these architectures in tasks such as image classification, object detection, and semantic segmentation. Vision Transformers (ViTs) have demonstrated impressive performance improvements over Convolutional Neural Networks (CNNs) due to their competitive modelling capabilities. However, these architectures demand massive computational resources which makes these models difficult to be deployed in the resource-constrained applications. Many solutions have been developed to combat this issue, such as compressive transformers and compression functions such as dilated convolution, min-max pooling, 1D convolution, etc. Model compression has recently attracted considerable research attention as a potential remedy. A number of model compression methods have been proposed in the literature such as weight quantization, weight multiplexing, pruning and Knowledge Distillation (KD). However, techniques like weight quantization, pruning and weight multiplexing typically involve complex pipelines for performing the compression. KD has been found to be a simple and much effective model compression technique that allows a relatively simple model to perform tasks almost as accurately as a complex model. This paper discusses various approaches based upon KD for effective compression of ViT models. The paper elucidates the role played by KD in reducing the computational and memory requirements of these models. The paper also presents the various challenges faced by ViTs that are yet to be resolved.

arXiv:2302.01530공식 코드 · Papers with Code

Revisiting Intermediate Layer Distillation for Compressing Language Models: An Overfitting Perspective

Knowledge distillation (KD) is a highly promising method for mitigating the computational problems of pre-trained language models (PLMs). Among various KD approaches, Intermediate Layer Distillation (ILD) has been a de facto standard KD method with its performance efficacy in the NLP field. In this paper, we find that existing ILD methods are prone to overfitting to training datasets, although these methods transfer more information than the original KD. Next, we present the simple observations to mitigate the overfitting of ILD: distilling only the last Transformer layer and conducting ILD on supplementary tasks. Based on our two findings, we propose a simple yet effective consistency-regularized ILD (CR-ILD), which prevents the student model from overfitting the training dataset. Substantial experiments on distilling BERT on the GLUE benchmark and several synthetic datasets demonstrate that our proposed ILD method outperforms other KD techniques. Our code is available at https://github.com/jongwooko/CR-ILD.