In real applications, different computation-resource devices need different-depth networks (e.g., ResNet-18/34/50) with high-accuracy. Usually, existing methods either design multiple networks and train them independently, or construct depth-level/width-level dynamic neural networks which is hard to prove the accuracy of each sub-net. In this article, we propose an elegant Depth-Level Dynamic Neural Network (DDNN) integrated different-depth sub-nets of similar architectures. To improve the generalization of sub-nets, we design the Embedded-Knowledge-Distillation (EKD) training mechanism for the DDNN to implement knowledge transfer from the teacher (full-net) to multiple students (sub-nets). Specifically, the Kullback-Leibler (KL) divergence is introduced to constrain the posterior class probability consistency between full-net and sub-nets, and self-attention distillation on the same resolution feature of different depth is addressed to drive more abundant feature representations of sub-nets. Thus, we can obtain multiple high-accuracy sub-nets simultaneously in a DDNN via the online knowledge distillation in each training iteration without extra computation cost. Extensive experiments on CIFAR-10/100, and ImageNet datasets demonstrate that sub-nets in DDNN with EKD training achieve better performance than individually training networks while preserving the original performance of full-nets.
相關內容
Knowledge distillation is a popular technique for training a small student network to emulate a larger teacher model, such as an ensemble of networks. We show that while knowledge distillation can improve student generalization, it does not typically work as it is commonly understood: there often remains a surprisingly large discrepancy between the predictive distributions of the teacher and the student, even in cases when the student has the capacity to perfectly match the teacher. We identify difficulties in optimization as a key reason for why the student is unable to match the teacher. We also show how the details of the dataset used for distillation play a role in how closely the student matches the teacher -- and that more closely matching the teacher paradoxically does not always lead to better student generalization.
Deep convolutional neural networks (CNNs) are often of sophisticated design with numerous convolutional layers and learnable parameters for the accuracy reason. To alleviate the expensive costs of deploying them on mobile devices, recent works have made huge efforts for excavating redundancy in pre-defined architectures. Nevertheless, the redundancy on the input resolution of modern CNNs has not been fully investigated, i.e., the resolution of input image is fixed. In this paper, we observe that the smallest resolution for accurately predicting the given image is different using the same neural network. To this end, we propose a novel dynamic-resolution network (DRNet) in which the resolution is determined dynamically based on each input sample. Thus, a resolution predictor with negligible computational costs is explored and optimized jointly with the desired network. In practice, the predictor learns the smallest resolution that can retain and even exceed the original recognition accuracy for each image. During the inference, each input image will be resized to its predicted resolution for minimizing the overall computation burden. We then conduct extensive experiments on several benchmark networks and datasets. The results show that our DRNet can be embedded in any off-the-shelf network architecture to obtain a considerable reduction in computational complexity. For instance, DRNet achieves similar performance with an about 34% computation reduction, while gains 1.4% accuracy increase with 10% computation reduction compared to the original ResNet-50 on ImageNet.
The existing solutions for object detection distillation rely on the availability of both a teacher model and ground-truth labels. We propose a new perspective to relax this constraint. In our framework, a student is first trained with pseudo labels generated by the teacher, and then fine-tuned using labeled data, if any available. Extensive experiments demonstrate improvements over existing object detection distillation algorithms. In addition, decoupling the teacher and ground-truth distillation in this framework provides interesting properties such: as 1) using unlabeled data to further improve the student's performance, 2) combining multiple teacher models of different architectures, even with different object categories, and 3) reducing the need for labeled data (with only 20% of COCO labels, this method achieves the same performance as the model trained on the entire set of labels). Furthermore, a by-product of this approach is the potential usage for domain adaptation. We verify these properties through extensive experiments.
Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) instance-wise dynamic models that process each instance with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions.
Knowledge distillation has demonstrated encouraging performances in deep model compression. Most existing approaches, however, require massive labeled data to accomplish the knowledge transfer, making the model compression a cumbersome and costly process. In this paper, we investigate the practical few-shot knowledge distillation scenario, where we assume only a few samples without human annotations are available for each category. To this end, we introduce a principled dual-stage distillation scheme tailored for few-shot data. In the first step, we graft the student blocks one by one onto the teacher, and learn the parameters of the grafted block intertwined with those of the other teacher blocks. In the second step, the trained student blocks are progressively connected and then together grafted onto the teacher network, allowing the learned student blocks to adapt themselves to each other and eventually replace the teacher network. Experiments demonstrate that our approach, with only a few unlabeled samples, achieves gratifying results on CIFAR10, CIFAR100, and ILSVRC-2012. On CIFAR10 and CIFAR100, our performances are even on par with those of knowledge distillation schemes that utilize the full datasets. The source code is available at //github.com/zju-vipa/NetGraft.
Adder Neural Networks (ANNs) which only contain additions bring us a new way of developing deep neural networks with low energy consumption. Unfortunately, there is an accuracy drop when replacing all convolution filters by adder filters. The main reason here is the optimization difficulty of ANNs using $\ell_1$-norm, in which the estimation of gradient in back propagation is inaccurate. In this paper, we present a novel method for further improving the performance of ANNs without increasing the trainable parameters via a progressive kernel based knowledge distillation (PKKD) method. A convolutional neural network (CNN) with the same architecture is simultaneously initialized and trained as a teacher network, features and weights of ANN and CNN will be transformed to a new space to eliminate the accuracy drop. The similarity is conducted in a higher-dimensional space to disentangle the difference of their distributions using a kernel based method. Finally, the desired ANN is learned based on the information from both the ground-truth and teacher, progressively. The effectiveness of the proposed method for learning ANN with higher performance is then well-verified on several benchmarks. For instance, the ANN-50 trained using the proposed PKKD method obtains a 76.8\% top-1 accuracy on ImageNet dataset, which is 0.6\% higher than that of the ResNet-50.
Knowledge graph (KG) embedding encodes the entities and relations from a KG into low-dimensional vector spaces to support various applications such as KG completion, question answering, and recommender systems. In real world, knowledge graphs (KGs) are dynamic and evolve over time with addition or deletion of triples. However, most existing models focus on embedding static KGs while neglecting dynamics. To adapt to the changes in a KG, these models need to be re-trained on the whole KG with a high time cost. In this paper, to tackle the aforementioned problem, we propose a new context-aware Dynamic Knowledge Graph Embedding (DKGE) method which supports the embedding learning in an online fashion. DKGE introduces two different representations (i.e., knowledge embedding and contextual element embedding) for each entity and each relation, in the joint modeling of entities and relations as well as their contexts, by employing two attentive graph convolutional networks, a gate strategy, and translation operations. This effectively helps limit the impacts of a KG update in certain regions, not in the entire graph, so that DKGE can rapidly acquire the updated KG embedding by a proposed online learning algorithm. Furthermore, DKGE can also learn KG embedding from scratch. Experiments on the tasks of link prediction and question answering in a dynamic environment demonstrate the effectiveness and efficiency of DKGE.
Knowledge distillation is typically conducted by training a small model (the student) to mimic a large and cumbersome model (the teacher). The idea is to compress the knowledge from the teacher by using its output probabilities as soft-labels to optimize the student. However, when the teacher is considerably large, there is no guarantee that the internal knowledge of the teacher will be transferred into the student; even if the student closely matches the soft-labels, its internal representations may be considerably different. This internal mismatch can undermine the generalization capabilities originally intended to be transferred from the teacher to the student. In this paper, we propose to distill the internal representations of a large model such as BERT into a simplified version of it. We formulate two ways to distill such representations and various algorithms to conduct the distillation. We experiment with datasets from the GLUE benchmark and consistently show that adding knowledge distillation from internal representations is a more powerful method than only using soft-label distillation.
Classifying large scale networks into several categories and distinguishing them according to their fine structures is of great importance with several applications in real life. However, most studies of complex networks focus on properties of a single network but seldom on classification, clustering, and comparison between different networks, in which the network is treated as a whole. Due to the non-Euclidean properties of the data, conventional methods can hardly be applied on networks directly. In this paper, we propose a novel framework of complex network classifier (CNC) by integrating network embedding and convolutional neural network to tackle the problem of network classification. By training the classifiers on synthetic complex network data and real international trade network data, we show CNC can not only classify networks in a high accuracy and robustness, it can also extract the features of the networks automatically.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191