PyTorchKR Papers

최신 논문

총 4,526편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1908.05161공식 코드 · Papers with Code

Scalable Attentive Sentence-Pair Modeling via Distilled Sentence Embedding

Recent state-of-the-art natural language understanding models, such as BERT and XLNet, score a pair of sentences (A and B) using multiple cross-attention operations - a process in which each word in sentence A attends to all words in sentence B and vice versa. As a result, computing the similarity between a query sentence and a set of candidate sentences, requires the propagation of all query-candidate sentence-pairs throughout a stack of cross-attention layers. This exhaustive process becomes computationally prohibitive when the number of candidate sentences is large. In contrast, sentence embedding techniques learn a sentence-to-vector mapping and compute the similarity between the sentence vectors via simple elementary operations. In this paper, we introduce Distilled Sentence Embedding (DSE) - a model that is based on knowledge distillation from cross-attentive models, focusing on sentence-pair tasks. The outline of DSE is as follows: Given a cross-attentive teacher model (e.g. a fine-tuned BERT), we train a sentence embedding based student model to reconstruct the sentence-pair scores obtained by the teacher model. We empirically demonstrate the effectiveness of DSE on five GLUE sentence-pair tasks. DSE significantly outperforms several ELMO variants and other sentence embedding methods, while accelerating computation of the query-candidate sentence-pairs similarities by several orders of magnitude, with an average relative degradation of 4.6% compared to BERT. Furthermore, we show that DSE produces sentence embeddings that reach state-of-the-art performance on universal sentence representation benchmarks. Our code is made publicly available at https://github.com/microsoft/Distilled-Sentence-Embedding.

arXiv:1908.04680

Effective Training of Convolutional Neural Networks with Low-bitwidth Weights and Activations

This paper tackles the problem of training a deep convolutional neural network of both low-bitwidth weights and activations. Optimizing a low-precision network is very challenging due to the non-differentiability of the quantizer, which may result in substantial accuracy loss. To address this, we propose three practical approaches, including (i) progressive quantization; (ii) stochastic precision; and (iii) joint knowledge distillation to improve the network training. First, for progressive quantization, we propose two schemes to progressively find good local minima. Specifically, we propose to first optimize a net with quantized weights and subsequently quantize activations. This is in contrast to the traditional methods which optimize them simultaneously. Furthermore, we propose a second progressive quantization scheme which gradually decreases the bit-width from high-precision to low-precision during training. Second, to alleviate the excessive training burden due to the multi-round training stages, we further propose a one-stage stochastic precision strategy to randomly sample and quantize sub-networks while keeping other parts in full-precision. Finally, we adopt a novel learning scheme to jointly train a full-precision model alongside the low-precision one. By doing so, the full-precision model provides hints to guide the low-precision model training and significantly improves the performance of the low-precision network. Extensive experiments on various datasets (e.g., CIFAR-100, ImageNet) show the effectiveness of the proposed methods.

arXiv:1908.01851

Self-Knowledge Distillation in Natural Language Processing

Since deep learning became a key player in natural language processing (NLP), many deep learning models have been showing remarkable performances in a variety of NLP tasks, and in some cases, they are even outperforming humans. Such high performance can be explained by efficient knowledge representation of deep learning models. While many methods have been proposed to learn more efficient representation, knowledge distillation from pretrained deep networks suggest that we can use more information from the soft target probability to train other neural networks. In this paper, we propose a new knowledge distillation method self-knowledge distillation, based on the soft target probabilities of the training model itself, where multimode information is distilled from the word embedding space right below the softmax layer. Due to the time complexity, our method approximates the soft target probabilities. In experiments, we applied the proposed method to two different and fundamental NLP tasks: language model and neural machine translation. The experiment results show that our proposed method improves performance on the tasks.

arXiv:1908.01581

Knowledge Consistency between Neural Networks and Beyond

This paper aims to analyze knowledge consistency between pre-trained deep neural networks. We propose a generic definition for knowledge consistency between neural networks at different fuzziness levels. A task-agnostic method is designed to disentangle feature components, which represent the consistent knowledge, from raw intermediate-layer features of each neural network. As a generic tool, our method can be broadly used for different applications. In preliminary experiments, we have used knowledge consistency as a tool to diagnose representations of neural networks. Knowledge consistency provides new insights to explain the success of existing deep-learning techniques, such as knowledge distillation and network compression. More crucially, knowledge consistency can also be used to refine pre-trained networks and boost performance.

arXiv:1908.00821공식 코드 · Papers with Code

Learning Lightweight Lane Detection CNNs by Self Attention Distillation

Training deep models for lane detection is challenging due to the very subtle and sparse supervisory signals inherent in lane annotations. Without learning from much richer context, these models often fail in challenging scenarios, e.g., severe occlusion, ambiguous lanes, and poor lighting conditions. In this paper, we present a novel knowledge distillation approach, i.e., Self Attention Distillation (SAD), which allows a model to learn from itself and gains substantial improvement without any additional supervision or labels. Specifically, we observe that attention maps extracted from a model trained to a reasonable level would encode rich contextual information. The valuable contextual information can be used as a form of 'free' supervision for further representation learning through performing topdown and layer-wise attention distillation within the network itself. SAD can be easily incorporated in any feedforward convolutional neural networks (CNN) and does not increase the inference time. We validate SAD on three popular lane detection benchmarks (TuSimple, CULane and BDD100K) using lightweight models such as ENet, ResNet-18 and ResNet-34. The lightest model, ENet-SAD, performs comparatively or even surpasses existing algorithms. Notably, ENet-SAD has 20 x fewer parameters and runs 10 x faster compared to the state-of-the-art SCNN, while still achieving compelling performance in all benchmarks. Our code is available at https://github.com/cardwing/Codes-for-Lane-Detection.

arXiv:1907.12926

Distill-to-Label: Weakly Supervised Instance Labeling Using Knowledge Distillation

Weakly supervised instance labeling using only image-level labels, in lieu of expensive fine-grained pixel annotations, is crucial in several applications including medical image analysis. In contrast to conventional instance segmentation scenarios in computer vision, the problems that we consider are characterized by a small number of training images and non-local patterns that lead to the diagnosis. In this paper, we explore the use of multiple instance learning (MIL) to design an instance label generator under this weakly supervised setting. Motivated by the observation that an MIL model can handle bags of varying sizes, we propose to repurpose an MIL model originally trained for bag-level classification to produce reliable predictions for single instances, i.e., bags of size $1$. To this end, we introduce a novel regularization strategy based on virtual adversarial training for improving MIL training, and subsequently develop a knowledge distillation technique for repurposing the trained MIL model. Using empirical studies on colon cancer and breast cancer detection from histopathological images, we show that the proposed approach produces high-quality instance-level prediction and significantly outperforms state-of-the MIL methods.

arXiv:1907.10107

Lifelong GAN: Continual Learning for Conditional Image Generation

Lifelong learning is challenging for deep neural networks due to their susceptibility to catastrophic forgetting. Catastrophic forgetting occurs when a trained network is not able to maintain its ability to accomplish previously learned tasks when it is trained to perform new tasks. We study the problem of lifelong learning for generative models, extending a trained network to new conditional generation tasks without forgetting previous tasks, while assuming access to the training data for the current task only. In contrast to state-of-the-art memory replay based approaches which are limited to label-conditioned image generation tasks, a more generic framework for continual learning of generative models under different conditional image generation settings is proposed in this paper. Lifelong GAN employs knowledge distillation to transfer learned knowledge from previous networks to the new network. This makes it possible to perform image-conditioned generation tasks in a lifelong learning setting. We validate Lifelong GAN for both image-conditioned and label-conditioned generation tasks, and provide qualitative and quantitative results to show the generality and effectiveness of our method.

arXiv:1907.10586

Distilled Siamese Networks for Visual Tracking

In recent years, Siamese network based trackers have significantly advanced the state-of-the-art in real-time tracking. Despite their success, Siamese trackers tend to suffer from high memory costs, which restrict their applicability to mobile devices with tight memory budgets. To address this issue, we propose a distilled Siamese tracking framework to learn small, fast and accurate trackers (students), which capture critical knowledge from large Siamese trackers (teachers) by a teacher-students knowledge distillation model. This model is intuitively inspired by the one teacher vs. multiple students learning method typically employed in schools. In particular, our model contains a single teacher-student distillation module and a student-student knowledge sharing mechanism. The former is designed using a tracking-specific distillation strategy to transfer knowledge from a teacher to students. The latter is utilized for mutual learning between students to enable in-depth knowledge understanding. Extensive empirical evaluations on several popular Siamese trackers demonstrate the generality and effectiveness of our framework. Moreover, the results on five tracking benchmarks show that the proposed distilled trackers achieve compression rates of up to 18$\times$ and frame-rates of $265$ FPS, while obtaining comparable tracking accuracy compared to base models.

arXiv:1907.09682

Similarity-Preserving Knowledge Distillation

Knowledge distillation is a widely applicable technique for training a student neural network under the guidance of a trained teacher network. For example, in neural network compression, a high-capacity teacher is distilled to train a compact student; in privileged learning, a teacher trained with privileged data is distilled to train a student without access to that data. The distillation loss determines how a teacher's knowledge is captured and transferred to the student. In this paper, we propose a new form of knowledge distillation loss that is inspired by the observation that semantically similar inputs tend to elicit similar activation patterns in a trained network. Similarity-preserving knowledge distillation guides the training of a student network such that input pairs that produce similar (dissimilar) activations in the teacher network produce similar (dissimilar) activations in the student network. In contrast to previous distillation methods, the student is not required to mimic the representation space of the teacher, but rather to preserve the pairwise similarities in its own representation space. Experiments on three public datasets demonstrate the potential of our approach.

arXiv:1907.09643공식 코드 · Papers with Code

Highlight Every Step: Knowledge Distillation via Collaborative Teaching

High storage and computational costs obstruct deep neural networks to be deployed on resource-constrained devices. Knowledge distillation aims to train a compact student network by transferring knowledge from a larger pre-trained teacher model. However, most existing methods on knowledge distillation ignore the valuable information among training process associated with training results. In this paper, we provide a new Collaborative Teaching Knowledge Distillation (CTKD) strategy which employs two special teachers. Specifically, one teacher trained from scratch (i.e., scratch teacher) assists the student step by step using its temporary outputs. It forces the student to approach the optimal path towards the final logits with high accuracy. The other pre-trained teacher (i.e., expert teacher) guides the student to focus on a critical region which is more useful for the task. The combination of the knowledge from two special teachers can significantly improve the performance of the student network in knowledge distillation. The results of experiments on CIFAR-10, CIFAR-100, SVHN and Tiny ImageNet datasets verify that the proposed knowledge distillation method is efficient and achieves state-of-the-art performance.

arXiv:1907.09831공식 코드 · Papers with Code

Real-Time Correlation Tracking via Joint Model Compression and Transfer

Correlation filters (CF) have received considerable attention in visual tracking because of their computational efficiency. Leveraging deep features via off-the-shelf CNN models (e.g., VGG), CF trackers achieve state-of-the-art performance while consuming a large number of computing resources. This limits deep CF trackers to be deployed to many mobile platforms on which only a single-core CPU is available. In this paper, we propose to jointly compress and transfer off-the-shelf CNN models within a knowledge distillation framework. We formulate a CNN model pretrained from the image classification task as a teacher network, and distill this teacher network into a lightweight student network as the feature extractor to speed up CF trackers. In the distillation process, we propose a fidelity loss to enable the student network to maintain the representation capability of the teacher network. Meanwhile, we design a tracking loss to adapt the objective of the student network from object recognition to visual tracking. The distillation process is performed offline on multiple layers and adaptively updates the student network using a background-aware online learning scheme. Extensive experiments on five challenging datasets demonstrate that the lightweight student network accelerates the speed of state-of-the-art deep CF trackers to real-time on a single-core CPU while maintaining almost the same tracking accuracy.

arXiv:1907.06870공식 코드 · Papers with Code

Light Multi-segment Activation for Model Compression

Model compression has become necessary when applying neural networks (NN) into many real application tasks that can accept slightly-reduced model accuracy with strict tolerance to model complexity. Recently, Knowledge Distillation, which distills the knowledge from well-trained and highly complex teacher model into a compact student model, has been widely used for model compression. However, under the strict requirement on the resource cost, it is quite challenging to achieve comparable performance with the teacher model, essentially due to the drastically-reduced expressiveness ability of the compact student model. Inspired by the nature of the expressiveness ability in Neural Networks, we propose to use multi-segment activation, which can significantly improve the expressiveness ability with very little cost, in the compact student model. Specifically, we propose a highly efficient multi-segment activation, called Light Multi-segment Activation (LMA), which can rapidly produce multiple linear regions with very few parameters by leveraging the statistical information. With using LMA, the compact student model is capable of achieving much better performance effectively and efficiently, than the ReLU-equipped one with same model scale. Furthermore, the proposed method is compatible with other model compression techniques, such as quantization, which means they can be used jointly for better compression performance. Experiments on state-of-the-art NN architectures over the real-world tasks demonstrate the effectiveness and extensibility of the LMA.

arXiv:1907.06017

Learn Spelling from Teachers: Transferring Knowledge from Language Models to Sequence-to-Sequence Speech Recognition

Integrating an external language model into a sequence-to-sequence speech recognition system is non-trivial. Previous works utilize linear interpolation or a fusion network to integrate external language models. However, these approaches introduce external components, and increase decoding computation. In this paper, we instead propose a knowledge distillation based training approach to integrating external language models into a sequence-to-sequence model. A recurrent neural network language model, which is trained on large scale external text, generates soft labels to guide the sequence-to-sequence model training. Thus, the language model plays the role of the teacher. This approach does not add any external component to the sequence-to-sequence model during testing. And this approach is flexible to be combined with shallow fusion technique together for decoding. The experiments are conducted on public Chinese datasets AISHELL-1 and CLMAD. Our approach achieves a character error rate of 9.3%, which is relatively reduced by 18.42% compared with the vanilla sequence-to-sequence model.

arXiv:1907.04829공식 코드 · Papers with Code

BAM! Born-Again Multi-Task Networks for Natural Language Understanding

It can be challenging to train multi-task neural networks that outperform or even match their single-task counterparts. To help address this, we propose using knowledge distillation where single-task models teach a multi-task model. We enhance this training with teacher annealing, a novel method that gradually transitions the model from distillation to supervised learning, helping the multi-task model surpass its single-task teachers. We evaluate our approach by multi-task fine-tuning BERT on the GLUE benchmark. Our method consistently improves over standard single-task and multi-task training.

arXiv:1907.02226공식 코드 · Papers with Code

Graph-based Knowledge Distillation by Multi-head Attention Network

Knowledge distillation (KD) is a technique to derive optimal performance from a small student network (SN) by distilling knowledge of a large teacher network (TN) and transferring the distilled knowledge to the small SN. Since a role of convolutional neural network (CNN) in KD is to embed a dataset so as to perform a given task well, it is very important to acquire knowledge that considers intra-data relations. Conventional KD methods have concentrated on distilling knowledge in data units. To our knowledge, any KD methods for distilling information in dataset units have not yet been proposed. Therefore, this paper proposes a novel method that enables distillation of dataset-based knowledge from the TN using an attention network. The knowledge of the embedding procedure of the TN is distilled to graph by multi-head attention (MHA), and multi-task learning is performed to give relational inductive bias to the SN. The MHA can provide clear information about the source dataset, which can greatly improves the performance of the SN. Experimental results show that the proposed method is 7.05% higher than the SN alone for CIFAR100, which is 2.46% higher than the state-of-the-art.

arXiv:1907.00873

Compression of Acoustic Event Detection Models With Quantized Distillation

Acoustic Event Detection (AED), aiming at detecting categories of events based on audio signals, has found application in many intelligent systems. Recently deep neural network significantly advances this field and reduces detection errors to a large scale. However how to efficiently execute deep models in AED has received much less attention. Meanwhile state-of-the-art AED models are based on large deep models, which are computational demanding and challenging to deploy on devices with constrained computational resources. In this paper, we present a simple yet effective compression approach which jointly leverages knowledge distillation and quantization to compress larger network (teacher model) into compact network (student model). Experimental results show proposed technique not only lowers error rate of original compact network by 15% through distillation but also further reduces its model size to a large extent (2% of teacher, 12% of full-precision student) through quantization.

arXiv:1906.12181

Reconstructing Perceived Images from Brain Activity by Visually-guided Cognitive Representation and Adversarial Learning

Reconstructing visual stimulus (image) only from human brain activity measured with functional Magnetic Resonance Imaging (fMRI) is a significant and meaningful task in Human-AI collaboration. However, the inconsistent distribution and representation between fMRI signals and visual images cause the heterogeneity gap. Moreover, the fMRI data is often extremely high-dimensional and contains a lot of visually-irrelevant information. Existing methods generally suffer from these issues so that a satisfactory reconstruction is still challenging. In this paper, we show that it is possible to overcome these challenges by learning visually-guided cognitive latent representations from the fMRI signals, and inversely decoding them to the image stimuli. The resulting framework is called Dual-Variational Autoencoder/ Generative Adversarial Network (D-VAE/GAN), which combines the advantages of adversarial representation learning with knowledge distillation. In addition, we introduce a novel three-stage learning approach which enables the (cognitive) encoder to gradually distill useful knowledge from the paired (visual) encoder during the learning process. Extensive experimental results on both artificial and natural images have demonstrated that our method could achieve surprisingly good results and outperform all other alternatives.

arXiv:1906.10834

Essence Knowledge Distillation for Speech Recognition

It is well known that a speech recognition system that combines multiple acoustic models trained on the same data significantly outperforms a single-model system. Unfortunately, real time speech recognition using a whole ensemble of models is too computationally expensive. In this paper, we propose to distill the knowledge of essence in an ensemble of models (i.e. the teacher model) to a single model (i.e. the student model) that needs much less computation to deploy. Previously, all the soften outputs of the teacher model are used to optimize the student model. We argue that not all the outputs of the ensemble are necessary to be distilled. Some of the outputs may even contain noisy information that is useless or even harmful to the training of the student model. In addition, we propose to train the student model with a multitask learning approach by utilizing both the soften outputs of the teacher model and the correct hard labels. The proposed method achieves some surprising results on the Switchboard data set. When the student model is trained together with the correct labels and the essence knowledge from the teacher model, it not only significantly outperforms another single model with the same architecture that is trained only with the correct labels, but also consistently outperforms the teacher model that is used to generate the soft labels.

arXiv:1906.09308공식 코드 · Papers with Code

Approximating Interactive Human Evaluation with Self-Play for Open-Domain Dialog Systems

Building an open-domain conversational agent is a challenging problem. Current evaluation methods, mostly post-hoc judgments of static conversation, do not capture conversation quality in a realistic interactive context. In this paper, we investigate interactive human evaluation and provide evidence for its necessity; we then introduce a novel, model-agnostic, and dataset-agnostic method to approximate it. In particular, we propose a self-play scenario where the dialog system talks to itself and we calculate a combination of proxies such as sentiment and semantic coherence on the conversation trajectory. We show that this metric is capable of capturing the human-rated quality of a dialog model better than any automated metric known to-date, achieving a significant Pearson correlation (r>.7, p<.05). To investigate the strengths of this novel metric and interactive evaluation in comparison to state-of-the-art metrics and human evaluation of static conversations, we perform extended experiments with a set of models, including several that make novel improvements to recent hierarchical dialog generation architectures through sentiment and semantic knowledge distillation on the utterance level. Finally, we open-source the interactive evaluation platform we built and the dataset we collected to allow researchers to efficiently deploy and evaluate dialog models.

arXiv:1906.08467

GAN-Knowledge Distillation for one-stage Object Detection

Convolutional neural networks have a significant improvement in the accuracy of Object detection. As convolutional neural networks become deeper, the accuracy of detection is also obviously improved, and more floating-point calculations are needed. Many researchers use the knowledge distillation method to improve the accuracy of student networks by transferring knowledge from a deeper and larger teachers network to a small student network, in object detection. Most methods of knowledge distillation need to designed complex cost functions and they are aimed at the two-stage object detection algorithm. This paper proposes a clean and effective knowledge distillation method for the one-stage object detection. The feature maps generated by teacher network and student network are used as true samples and fake samples respectively, and generate adversarial training for both to improve the performance of the student network in one-stage object detection.