PyTorchKR Papers

최신 논문

총 4,526편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2212.06552

Domain Adaptation for Dense Retrieval through Self-Supervision by Pseudo-Relevance Labeling

Although neural information retrieval has witnessed great improvements, recent works showed that the generalization ability of dense retrieval models on target domains with different distributions is limited, which contrasts with the results obtained with interaction-based models. To address this issue, researchers have resorted to adversarial learning and query generation approaches; both approaches nevertheless resulted in limited improvements. In this paper, we propose to use a self-supervision approach in which pseudo-relevance labels are automatically generated on the target domain. To do so, we first use the standard BM25 model on the target domain to obtain a first ranking of documents, and then use the interaction-based model T53B to re-rank top documents. We further combine this approach with knowledge distillation relying on an interaction-based teacher model trained on the source domain. Our experiments reveal that pseudo-relevance labeling using T53B and the MiniLM teacher performs on average better than other approaches and helps improve the state-of-the-art query generation approach GPL when it is fine-tuned on the pseudo-relevance labeled data.

arXiv:2212.05699

Multimodal Matching-aware Co-attention Networks with Mutual Knowledge Distillation for Fake News Detection

Fake news often involves multimedia information such as text and image to mislead readers, proliferating and expanding its influence. Most existing fake news detection methods apply the co-attention mechanism to fuse multimodal features while ignoring the consistency of image and text in co-attention. In this paper, we propose multimodal matching-aware co-attention networks with mutual knowledge distillation for improving fake news detection. Specifically, we design an image-text matching-aware co-attention mechanism which captures the alignment of image and text for better multimodal fusion. The image-text matching representation can be obtained via a vision-language pre-trained model. Additionally, based on the designed image-text matching-aware co-attention mechanism, we propose to build two co-attention networks respectively centered on text and image for mutual knowledge distillation to improve fake news detection. Extensive experiments on three benchmark datasets demonstrate that our proposed model achieves state-of-the-art performance on multimodal fake news detection.

arXiv:2212.05225공식 코드 · Papers with Code

LEAD: Liberal Feature-based Distillation for Dense Retrieval

Knowledge distillation is often used to transfer knowledge from a strong teacher model to a relatively weak student model. Traditional methods include response-based methods and feature-based methods. Response-based methods are widely used but suffer from lower upper limits of performance due to their ignorance of intermediate signals, while feature-based methods have constraints on vocabularies, tokenizers and model architectures. In this paper, we propose a liberal feature-based distillation method (LEAD). LEAD aligns the distribution between the intermediate layers of teacher model and student model, which is effective, extendable, portable and has no requirements on vocabularies, tokenizers, or model architectures. Extensive experiments show the effectiveness of LEAD on widely-used benchmarks, including MS MARCO Passage Ranking, TREC 2019 DL Track, MS MARCO Document Ranking and TREC 2020 DL Track. Our code is available in https://github.com/microsoft/SimXNS/tree/main/LEAD.

arXiv:2212.05998

Continuation KD: Improved Knowledge Distillation through the Lens of Continuation Optimization

Knowledge Distillation (KD) has been extensively used for natural language understanding (NLU) tasks to improve a small model's (a student) generalization by transferring the knowledge from a larger model (a teacher). Although KD methods achieve state-of-the-art performance in numerous settings, they suffer from several problems limiting their performance. It is shown in the literature that the capacity gap between the teacher and the student networks can make KD ineffective. Additionally, existing KD techniques do not mitigate the noise in the teacher's output: modeling the noisy behaviour of the teacher can distract the student from learning more useful features. We propose a new KD method that addresses these problems and facilitates the training compared to previous techniques. Inspired by continuation optimization, we design a training procedure that optimizes the highly non-convex KD objective by starting with the smoothed version of this objective and making it more complex as the training proceeds. Our method (Continuation-KD) achieves state-of-the-art performance across various compact architectures on NLU (GLUE benchmark) and computer vision tasks (CIFAR-10 and CIFAR-100).

arXiv:2212.05956

Improving Generalization of Pre-trained Language Models via Stochastic Weight Averaging

Knowledge Distillation (KD) is a commonly used technique for improving the generalization of compact Pre-trained Language Models (PLMs) on downstream tasks. However, such methods impose the additional burden of training a separate teacher model for every new dataset. Alternatively, one may directly work on the improvement of the optimization procedure of the compact model toward better generalization. Recent works observe that the flatness of the local minimum correlates well with better generalization. In this work, we adapt Stochastic Weight Averaging (SWA), a method encouraging convergence to a flatter minimum, to fine-tuning PLMs. We conduct extensive experiments on various NLP tasks (text classification, question answering, and generation) and different model architectures and demonstrate that our adaptation improves the generalization without extra computation cost. Moreover, we observe that this simple optimization technique is able to outperform the state-of-the-art KD methods for compact models.

arXiv:2212.05422

Teaching What You Should Teach: A Data-Based Distillation Method

In real teaching scenarios, an excellent teacher always teaches what he (or she) is good at but the student is not. This gives the student the best assistance in making up for his (or her) weaknesses and becoming a good one overall. Enlightened by this, we introduce the "Teaching what you Should Teach" strategy into a knowledge distillation framework, and propose a data-based distillation method named "TST" that searches for desirable augmented samples to assist in distilling more efficiently and rationally. To be specific, we design a neural network-based data augmentation module with priori bias, which assists in finding what meets the teacher's strengths but the student's weaknesses, by learning magnitudes and probabilities to generate suitable data samples. By training the data augmentation module and the generalized distillation paradigm in turn, a student model is learned with excellent generalization ability. To verify the effectiveness of our method, we conducted extensive comparative experiments on object recognition, detection, and segmentation tasks. The results on the CIFAR-10, ImageNet-1k, MS-COCO, and Cityscapes datasets demonstrate that our method achieves state-of-the-art performance on almost all teacher-student pairs. Furthermore, we conduct visualization studies to explore what magnitudes and probabilities are needed for the distillation process.

arXiv:2212.05330

Complete-to-Partial 4D Distillation for Self-Supervised Point Cloud Sequence Representation Learning

Recent work on 4D point cloud sequences has attracted a lot of attention. However, obtaining exhaustively labeled 4D datasets is often very expensive and laborious, so it is especially important to investigate how to utilize raw unlabeled data. However, most existing self-supervised point cloud representation learning methods only consider geometry from a static snapshot omitting the fact that sequential observations of dynamic scenes could reveal more comprehensive geometric details. And the video representation learning frameworks mostly model motion as image space flows, let alone being 3D-geometric-aware. To overcome such issues, this paper proposes a new 4D self-supervised pre-training method called Complete-to-Partial 4D Distillation. Our key idea is to formulate 4D self-supervised representation learning as a teacher-student knowledge distillation framework and let the student learn useful 4D representations with the guidance of the teacher. Experiments show that this approach significantly outperforms previous pre-training approaches on a wide range of 4D point cloud sequence understanding tasks including indoor and outdoor scenarios.

arXiv:2212.05973공식 코드 · Papers with Code

Towards Practical Plug-and-Play Diffusion Models

Diffusion-based generative models have achieved remarkable success in image generation. Their guidance formulation allows an external model to plug-and-play control the generation process for various tasks without finetuning the diffusion model. However, the direct use of publicly available off-the-shelf models for guidance fails due to their poor performance on noisy inputs. For that, the existing practice is to fine-tune the guidance models with labeled data corrupted with noises. In this paper, we argue that this practice has limitations in two aspects: (1) performing on inputs with extremely various noises is too hard for a single guidance model; (2) collecting labeled datasets hinders scaling up for various tasks. To tackle the limitations, we propose a novel strategy that leverages multiple experts where each expert is specialized in a particular noise range and guides the reverse process of the diffusion at its corresponding timesteps. However, as it is infeasible to manage multiple networks and utilize labeled data, we present a practical guidance framework termed Practical Plug-And-Play (PPAP), which leverages parameter-efficient fine-tuning and data-free knowledge transfer. We exhaustively conduct ImageNet class conditional generation experiments to show that our method can successfully guide diffusion with small trainable parameters and no labeled data. Finally, we show that image classifiers, depth estimators, and semantic segmentation models can guide publicly available GLIDE through our framework in a plug-and-play manner. Our code is available at https://github.com/riiid/PPAP.

arXiv:2212.04029

Occlusion-Robust FAU Recognition by Mining Latent Space of Masked Autoencoders

Facial action units (FAUs) are critical for fine-grained facial expression analysis. Although FAU detection has been actively studied using ideally high quality images, it was not thoroughly studied under heavily occluded conditions. In this paper, we propose the first occlusion-robust FAU recognition method to maintain FAU detection performance under heavy occlusions. Our novel approach takes advantage of rich information from the latent space of masked autoencoder (MAE) and transforms it into FAU features. Bypassing the occlusion reconstruction step, our model efficiently extracts FAU features of occluded faces by mining the latent space of a pretrained masked autoencoder. Both node and edge-level knowledge distillation are also employed to guide our model to find a mapping between latent space vectors and FAU features. Facial occlusion conditions, including random small patches and large blocks, are thoroughly studied. Experimental results on BP4D and DISFA datasets show that our method can achieve state-of-the-art performances under the studied facial occlusion, significantly outperforming existing baseline methods. In particular, even under heavy occlusion, the proposed method can achieve comparable performance as state-of-the-art methods under normal conditions.

arXiv:2212.03329공식 코드 · Papers with Code

Enhancing Low-Density EEG-Based Brain-Computer Interfaces with Similarity-Keeping Knowledge Distillation

Electroencephalogram (EEG) has been one of the common neuromonitoring modalities for real-world brain-computer interfaces (BCIs) because of its non-invasiveness, low cost, and high temporal resolution. Recently, light-weight and portable EEG wearable devices based on low-density montages have increased the convenience and usability of BCI applications. However, loss of EEG decoding performance is often inevitable due to reduced number of electrodes and coverage of scalp regions of a low-density EEG montage. To address this issue, we introduce knowledge distillation (KD), a learning mechanism developed for transferring knowledge/information between neural network models, to enhance the performance of low-density EEG decoding. Our framework includes a newly proposed similarity-keeping (SK) teacher-student KD scheme that encourages a low-density EEG student model to acquire the inter-sample similarity as in a pre-trained teacher model trained on high-density EEG data. The experimental results validate that our SK-KD framework consistently improves motor-imagery EEG decoding accuracy when number of electrodes deceases for the input EEG data. For both common low-density headphone-like and headband-like montages, our method outperforms state-of-the-art KD methods across various EEG decoding model architectures. As the first KD scheme developed for enhancing EEG decoding, we foresee the proposed SK-KD framework to facilitate the practicality of low-density EEG-based BCI in real-world applications.

arXiv:2212.02800

Life-long Learning for Multilingual Neural Machine Translation with Knowledge Distillation

A common scenario of Multilingual Neural Machine Translation (MNMT) is that each translation task arrives in a sequential manner, and the training data of previous tasks is unavailable. In this scenario, the current methods suffer heavily from catastrophic forgetting (CF). To alleviate the CF, we investigate knowledge distillation based life-long learning methods. Specifically, in one-tomany scenario, we propose a multilingual distillation method to make the new model (student) jointly learn multilingual output from old model (teacher) and new task. In many-to one scenario, we find that direct distillation faces the extreme partial distillation problem, and we propose two different methods to address it: pseudo input distillation and reverse teacher distillation. The experimental results on twelve translation tasks show that the proposed methods can better consolidate the previous knowledge and sharply alleviate the CF.

arXiv:2212.02969

Open World DETR: Transformer based Open World Object Detection

Open world object detection aims at detecting objects that are absent in the object classes of the training data as unknown objects without explicit supervision. Furthermore, the exact classes of the unknown objects must be identified without catastrophic forgetting of the previous known classes when the corresponding annotations of unknown objects are given incrementally. In this paper, we propose a two-stage training approach named Open World DETR for open world object detection based on Deformable DETR. In the first stage, we pre-train a model on the current annotated data to detect objects from the current known classes, and concurrently train an additional binary classifier to classify predictions into foreground or background classes. This helps the model to build an unbiased feature representations that can facilitate the detection of unknown classes in subsequent process. In the second stage, we fine-tune the class-specific components of the model with a multi-view self-labeling strategy and a consistency constraint. Furthermore, we alleviate catastrophic forgetting when the annotations of the unknown classes becomes available incrementally by using knowledge distillation and exemplar replay. Experimental results on PASCAL VOC and MS-COCO show that our proposed method outperforms other state-of-the-art open world object detection methods by a large margin.

arXiv:2212.02931

Leveraging Different Learning Styles for Improved Knowledge Distillation in Biomedical Imaging

Learning style refers to a type of training mechanism adopted by an individual to gain new knowledge. As suggested by the VARK model, humans have different learning preferences, like Visual (V), Auditory (A), Read/Write (R), and Kinesthetic (K), for acquiring and effectively processing information. Our work endeavors to leverage this concept of knowledge diversification to improve the performance of model compression techniques like Knowledge Distillation (KD) and Mutual Learning (ML). Consequently, we use a single-teacher and two-student network in a unified framework that not only allows for the transfer of knowledge from teacher to students (KD) but also encourages collaborative learning between students (ML). Unlike the conventional approach, where the teacher shares the same knowledge in the form of predictions or feature representations with the student network, our proposed approach employs a more diversified strategy by training one student with predictions and the other with feature maps from the teacher. We further extend this knowledge diversification by facilitating the exchange of predictions and feature maps between the two student networks, enriching their learning experiences. We have conducted comprehensive experiments with three benchmark datasets for both classification and segmentation tasks using two different network architecture combinations. These experimental results demonstrate that knowledge diversification in a combined KD and ML framework outperforms conventional KD or ML techniques (with similar network configuration) that only use predictions with an average improvement of 2%. Furthermore, consistent improvement in performance across different tasks, with various network architectures, and over state-of-the-art techniques establishes the robustness and generalizability of the proposed model

arXiv:2212.01543

The RoyalFlush System for the WMT 2022 Efficiency Task

This paper describes the submission of the RoyalFlush neural machine translation system for the WMT 2022 translation efficiency task. Unlike the commonly used autoregressive translation system, we adopted a two-stage translation paradigm called Hybrid Regression Translation (HRT) to combine the advantages of autoregressive and non-autoregressive translation. Specifically, HRT first autoregressively generates a discontinuous sequence (e.g., make a prediction every $k$ tokens, $k>1$) and then fills in all previously skipped tokens at once in a non-autoregressive manner. Thus, we can easily trade off the translation quality and speed by adjusting $k$. In addition, by integrating other modeling techniques (e.g., sequence-level knowledge distillation and deep-encoder-shallow-decoder layer allocation strategy) and a mass of engineering efforts, HRT improves 80\% inference speed and achieves equivalent translation performance with the same-capacity AT counterpart. Our fastest system reaches 6k+ words/second on the GPU latency setting, estimated to be about 3.1x faster than the last year's winner.

arXiv:2212.02379

Single image calibration using knowledge distillation approaches

Although recent deep learning-based calibration methods can predict extrinsic and intrinsic camera parameters from a single image, their generalization remains limited by the number and distribution of training data samples. The huge computational and space requirement prevents convolutional neural networks (CNNs) from being implemented in resource-constrained environments. This challenge motivated us to learn a CNN gradually, by training new data while maintaining performance on previously learned data. Our approach builds upon a CNN architecture to automatically estimate camera parameters (focal length, pitch, and roll) using different incremental learning strategies to preserve knowledge when updating the network for new data distributions. Precisely, we adapt four common incremental learning, namely: LwF , iCaRL, LU CIR, and BiC by modifying their loss functions to our regression problem. We evaluate on two datasets containing 299008 indoor and outdoor images. Experiment results were significant and indicated which method was better for the camera calibration estimation.

arXiv:2212.02196공식 코드 · Papers with Code

FedUKD: Federated UNet Model with Knowledge Distillation for Land Use Classification from Satellite and Street Views

Federated Deep Learning frameworks can be used strategically to monitor Land Use locally and infer environmental impacts globally. Distributed data from across the world would be needed to build a global model for Land Use classification. The need for a Federated approach in this application domain would be to avoid transfer of data from distributed locations and save network bandwidth to reduce communication cost. We use a Federated UNet model for Semantic Segmentation of satellite and street view images. The novelty of the proposed architecture is the integration of Knowledge Distillation to reduce communication cost and response time. The accuracy obtained was above 95% and we also brought in a significant model compression to over 17 times and 62 times for street View and satellite images respectively. Our proposed framework has the potential to be a game-changer in real-time tracking of climate change across the planet.

arXiv:2212.02057

DA-CIL: Towards Domain Adaptive Class-Incremental 3D Object Detection

Deep learning has achieved notable success in 3D object detection with the advent of large-scale point cloud datasets. However, severe performance degradation in the past trained classes, i.e., catastrophic forgetting, still remains a critical issue for real-world deployment when the number of classes is unknown or may vary. Moreover, existing 3D class-incremental detection methods are developed for the single-domain scenario, which fail when encountering domain shift caused by different datasets, varying environments, etc. In this paper, we identify the unexplored yet valuable scenario, i.e., class-incremental learning under domain shift, and propose a novel 3D domain adaptive class-incremental object detection framework, DA-CIL, in which we design a novel dual-domain copy-paste augmentation method to construct multiple augmented domains for diversifying training distributions, thereby facilitating gradual domain adaptation. Then, multi-level consistency is explored to facilitate dual-teacher knowledge distillation from different domains for domain adaptive class-incremental learning. Extensive experiments on various datasets demonstrate the effectiveness of the proposed method over baselines in the domain adaptive class-incremental learning scenario.

arXiv:2212.01012

Injecting Spatial Information for Monaural Speech Enhancement via Knowledge Distillation

Monaural speech enhancement (SE) provides a versatile and cost-effective approach to SE tasks by utilizing recordings from a single microphone. However, the monaural SE lags performance behind multi-channel SE as the monaural SE methods are unable to extract spatial information from one-channel recordings, which greatly limits their application scenarios. To address this issue, we inject spatial information into the monaural SE model and propose a knowledge distillation strategy to enable the monaural SE model to learn binaural speech features from the binaural SE model, which makes monaural SE model possible to reconstruct higher intelligibility and quality enhanced speeches under low signal-to-noise ratio (SNR) conditions. Extensive experiments show that our proposed monaural SE model by injecting spatial information via knowledge distillation achieves favorable performance against other monaural SE models with fewer parameters.

arXiv:2212.01188공식 코드 · Papers with Code

Improving Simultaneous Machine Translation with Monolingual Data

Simultaneous machine translation (SiMT) is usually done via sequence-level knowledge distillation (Seq-KD) from a full-sentence neural machine translation (NMT) model. However, there is still a significant performance gap between NMT and SiMT. In this work, we propose to leverage monolingual data to improve SiMT, which trains a SiMT student on the combination of bilingual data and external monolingual data distilled by Seq-KD. Preliminary experiments on En-Zh and En-Ja news domain corpora demonstrate that monolingual data can significantly improve translation quality (e.g., +3.15 BLEU on En-Zh). Inspired by the behavior of human simultaneous interpreters, we propose a novel monolingual sampling strategy for SiMT, considering both chunk length and monotonicity. Experimental results show that our sampling strategy consistently outperforms the random sampling strategy (and other conventional typical NMT monolingual sampling strategies) by avoiding the key problem of SiMT -- hallucination, and has better scalability. We achieve +0.72 BLEU improvements on average against random sampling on En-Zh and En-Ja. Data and codes can be found at https://github.com/hexuandeng/Mono4SiMT.