Explainability-based knowledge distillation

Knowledge distillation has been used to transfer knowledge learned by a sophisticated model (teacher) to a simpler model (student). This technique is widely used to compress model complexity. However, in most applications the compressed student model suffers from an accuracy gap with its teacher. We propose extracurricular learning, a novel knowledge distillation method, that bridges this gap by (1) modeling student and teacher output distributions; (2) sampling examples from an approximation to the underlying data distribution; and (3) matching student and teacher output distributions over this extended set including uncertain samples. We conduct rigorous evaluations on regression and classification tasks and show that compared to the standard knowledge distillation, extracurricular learning reduces the gap by 46% to 68%. This leads to major accuracy improvements compared to the empirical risk minimization-based training for various recent neural network architectures: 16% regression error reduction on the MPIIGaze dataset, +3.4% to +9.1% improvement in top-1 classification accuracy on the CIFAR100 dataset, and +2.9% top-1 improvement on the ImageNet dataset.

Breast cancer is the most common invasive cancer in women and the second main cause of cancer death in females, which can be classified benign or malignant. Research and prevention on breast cancer have attracted more concern of researchers in recent years. On the other hand, the development of data mining methods provides an effective way to extract more useful information from complex databases, and some prediction, classification, and clustering can be made according to the extracted information. The generic notion of knowledge distillation is that a network of higher capacity acts as a teacher and a network of lower capacity acts as a student. There are different pipelines of knowledge distillation known. However, previous work on knowledge distillation using label smoothing regularization produces experiments and results that break this general notion and prove that knowledge distillation also works when a student model distils a teacher model, i.e., reverse knowledge distillation. Not only this, but it is also proved that a poorly trained teacher model trains a student model to reach equivalent results. Building on the ideas from those works, we propose a novel bilateral knowledge distillation regime that enables multiple interactions between teacher and student models, i.e., teaching and distilling each other, eventually improving each other's performance and evaluating our results on BACH histopathology image dataset on breast cancer. The pretrained ResNeXt29 and MobileNetV2 models which are already tested on ImageNet dataset are used for “transfer learning” in our dataset, and we obtain a final accuracy of more than 96% using this novel approach of bilateral KD.

Knowledge distillation (KD) is an established paradigm for transferring privileged knowledge from a cumbersome model to a more lightweight and efficient one. In recent years, logit-based KD methods are quickly catching up in performance with their feature-based counterparts. However, previous research has pointed out that logit-based methods are still fundamentally limited by two major issues in their training process, namely overconfident teacher and confirmation bias. Inspired by the success of cross-view learning in fields such as semi-supervised learning, in this work we introduce within-view and cross-view regularisations to standard logit-based distillation frameworks to combat the above cruxes. We also perform confidence-based soft label mining to improve the quality of distilling signals from the teacher, which further mitigates the confirmation bias problem. Despite its apparent simplicity, the proposed Consistency-Regularisation-based Logit Distillation (CRLD) significantly boosts student learning, setting new state-of-the-art results on the standard CIFAR-100, Tiny-ImageNet, and ImageNet datasets across a diversity of teacher and student architectures, whilst introducing no extra network parameters. Orthogonal to on-going logit-based distillation research, our method enjoys excellent generalisation properties and, without bells and whistles, boosts the performance of various existing approaches by considerable margins.

Maximizing discrimination capability of knowledge distillation with energy function

Despite the fact that deep neural networks are powerful models and achieve appealing results on many tasks, they are too large to be deployed on edge devices like smartphones or embedded sensor nodes. There have been efforts to compress these networks, and a popular method is knowledge distillation, where a large (teacher) pre-trained network is used to train a smaller (student) network. However, in this paper, we show that the student network performance degrades when the gap between student and teacher is large. Given a fixed student network, one cannot employ an arbitrarily large teacher, or in other words, a teacher can effectively transfer its knowledge to students up to a certain size, not smaller. To alleviate this shortcoming, we introduce multi-step knowledge distillation, which employs an intermediate-sized network (teacher assistant) to bridge the gap between the student and the teacher. Moreover, we study the effect of teacher assistant size and extend the framework to multi-step distillation. Theoretical analysis and extensive experiments on CIFAR-10,100 and ImageNet datasets and on CNN and ResNet architectures substantiate the effectiveness of our proposed approach.

Knowledge distillation aims to distill knowledge from teacher networks to train student networks. Distilling intermediate features has attracted much attention in recent years as it can be flexibly applied in various fields such as image classification, object detection and semantic segmentation. A critical obstacle of feature-based knowledge distillation is the dimension gap between the intermediate features of teacher and student, and plenty of methods have been proposed to resolve this problem. However, these works usually implement feature uniformization in an unsupervised way, lacking guidance to help the student network learn meaningful mapping functions in the uniformization process. Moreover, the dimension uniformization process of the student and teacher network is usually not equivalent as the mapping functions are different. To this end, some factors of the feature are discarded during parametric feature alignment, or some factors are blended in some non-parametric operations. In this paper, we propose a novel semantic-aware knowledge distillation scheme to solve these problems. We build a standalone feature-based classification branch to extract semantic-aware knowledge for better guiding the learning process of the student network. In addition, to avoid the inequivalence of feature uniformization between teacher and student, we design a novel parameter-free self-attention operation that can convert features of different dimensions into vectors of the same length. Experimental results show that the proposed knowledge distillation scheme outperforms existing feature-based distillation methods on the widely used CIFAR-100 and CINIC-10 datasets.

Deep architecture can improve performance of neural networks whereas it increases the computational complexity. Compressing networks is the key to solve this problem. The framework Knowledge Distilling (KD) compresses cumbersome networks well. It improved mimic learning, enabling knowledge to be transferred from cumbersome networks to compressed networks without constraint of architectures. Inspired by AlphaGo Zero, this paper proposed an algorithm combining KD with reinforcement learning to compress networks on changing datasets. In this algorithm, the compressed networks interact with the environment made by KD to produce datasets that are appropriate w.r.t the model. Monte Carlo Tree Search (MCTS) of AlphaGo Zero is used to produce the datasets by making a trade-off between the prediction of compressed networks and the knowledge. In experiments, the algorithm proved to be effective in compressing networks by training ResNet on CIFAR datasets, with mean squared error as the object function.

Compared to traditional learning from scratch, knowledge distillation sometimes makes the DNN achieve superior performance. In this paper, we provide a new perspective to explain the success of knowledge distillation based on the information theory, i.e., quantifying knowledge points encoded in intermediate layers of a DNN for classification. To this end, we consider the signal processing in a DNN as a layer-wise process of discarding information. A knowledge point is referred to as an input unit, the information of which is discarded much less than that of other input units. Thus, we propose three hypotheses for knowledge distillation based on the quantification of knowledge points. 1. The DNN learning from knowledge distillation encodes more knowledge points than the DNN learning from scratch. 2. Knowledge distillation makes the DNN more likely to learn different knowledge points simultaneously. In comparison, the DNN learning from scratch tends to encode various knowledge points sequentially. 3. The DNN learning from knowledge distillation is often more stably optimized than the DNN learning from scratch. To verify the above hypotheses, we design three types of metrics with annotations of foreground objects to analyze feature representations of the DNN, i.e., the quantity and the quality of knowledge points, the learning speed of different knowledge points, and the stability of optimization directions. In experiments, we diagnosed various DNNs on different classification tasks, including image classification, 3D point cloud classification, binary sentiment classification, and question answering, which verified the above hypotheses.

Hyperspectral image (HSI) classification models suffer from a phenomenon known as catastrophic forgetting, which refers to the sharp decline in performance on previously learned tasks after learning a new one when continuously acquiring new knowledge from a sequence of tasks. In recent years, some lifelong learning approaches have been proposed for HSI classification. Despite some progress, the challenge of catastrophic forgetting in lifelong learning remains significant and unresolved. In this article, we propose a novel lifelong learning framework for HSI classification, which is based on exemplar replay and cross-spectral–spatial feature knowledge distillation (KD), termed HyperKD. Specifically, the proposed framework incorporates a min-max cross-selection (MMCS) module tailored to HSI characteristics with a cross-spectral-spatial knowledge distillation (CSSKD) module. The MMCS module selects the most representative or diverse samples as exemplars from previous tasks for replay. Additionally, the CSSKD module not only transfers the prediction logit distribution from the previous network to the current network but also transfers the spectral-spatial feature distribution via cross-network KD, without directly assessing the similarity of feature distributions, thereby retaining more knowledge and mitigating forgetting. Through experiments conducted on a series of tasks, including the Pavia, Indian Pines, Salinas, and Houston datasets, our approach demonstrates superior performance compared to previous lifelong learning methods for HSI classification, effectively mitigating catastrophic forgetting. The code implementation of our approach will be publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/lzlsxs/hyperkd</uri>.

Knowledge distillation is a widely used neural network model compression technique. In general, the knowledge distillation transfer the knowledge from a large pre-trained teacher network with superior performance to a small student network enables the student network to achieve better performance. This paper proposes a triplet knowledge distillation framework (abbreviated as TKD), which introduces a smaller assistant network into the knowledge distillation structure. The performance of the assistant network is lower than that of the student network. During the training of the TKD, by minimizing the Mean Squared Error(MSE) loss function, the output of the student network will closer to the output of the teacher network and further from that of the assistant network. Therefore, the student network can learn more expressive knowledge from the teacher network while throwing away mistaken knowledge in the assistant network. Finally, the student network achieves a surprising performance even superior to the teacher network. We have demonstrated the effectiveness of TKD by extensive experiments on benchmark datasets(CIFAR-10, CIFAR-100, SVHN, STL-10). When using VGGNet as an experimental model, the student network VGGNet13 achieving 94.29%, 75.30%, 95.53%, and 87.61% accuracy on the CIFAR-10, CIFAR-100, SVHN, and STL-10 datasets, improved by 1.24%, 2.81%, 0.40%, and 2.32%, respectively.

Weakly supervised methods, such as class activation maps (CAM) based, have been applied to achieve bleeding segmentation with low annotation efforts in Wireless Capsule Endoscopy (WCE) images. However, the CAM labels tend to be extremely noisy, and there is an irreparable gap between CAM labels and ground truths for medical images. This paper proposes a new Discrepancy-basEd Active Learning (DEAL) approach to bridge the gap between CAMs and ground truths with a few annotations. Specifically, to liberate labor, we design a novel discrepancy decoder model and a CAMPUS (CAM, Pseudo-label and groUnd-truth Selection) criterion to replace the noisy CAMs with accurate model predictions and a few human labels. The discrepancy decoder model is trained with a unique scheme to generate standard, coarse and fine predictions. And the CAMPUS criterion is proposed to predict the gaps between CAMs and ground truths based on model divergence and CAM divergence. We evaluate our method on the WCE dataset and results show that our method outperforms the state-of-the-art active learning methods and reaches comparable performance to those trained with full annotated datasets with only 10% of the training data labeled. The source code is available at https://github.com/baifanxxx/DEAL.KeywordsActive learningSegmentationWCE images

Knowledge distillation (KD) aims to distill the knowledge from the teacher (larger) to the student (smaller) model via soft-label for the efficient neural network. In general, the performance of a model is determined by accuracy, which is measured with labels. However, existing KD approaches usually use the teacher with its original distribution, neglecting the potential of incorrect prediction. This may contradict the motivation of hard-label learning through cross-entropy loss, which may lead to sub-optimal knowledge distillation on certain samples. To address this issue, we propose a novel logit processing scheme via a sorting mechanism. Specifically, our method has a two-fold goal: (1) fixing the incorrect prediction of the teacher based on the labels and (2) reordering the distribution in a natural way according to priority rank at once. As an easy-to-use, plug-andplay pre-processing, our sort method can be effectively applied to existing logit-based KD methods. Extensive experiments on the CIFAR-100 and ImageNet datasets demonstrate the effectiveness of our method.

Improving adversarial robustness using knowledge distillation guided by attention information bottleneck

Convolutional neural networks (CNNs) achieve prevailing results in segmentation tasks nowadays and represent the state-of-the-art for image-based analysis. However, the under- standing of the accurate decision-making process of a CNN is rather unknown. The research area of explainable artificial intelligence (xAI) primarily revolves around understanding and interpreting this black-box behavior. One way of interpreting a CNN is the use of class ac- tivation maps (CAMs) that represent heatmaps to indicate the importance of image areas for the prediction of the CNN. For classification tasks, a variety of CAM algorithms exist. But for segmentation tasks, only one CAM algorithm for the interpretation of the output of a CNN exist. We propose a transfer between existing classification- and segmentation- based methods for more detailed, explainable, and consistent results which show salient pixels in semantic segmentation tasks. The resulting Seg-HiRes-Grad CAM is an exten- sion of the segmentation-based Seg-Grad CAM with the transfer to the classification-based HiRes CAM. Our method improves the previously-mentioned existing segmentation-based method by adjusting it to recently published classification-based methods. Especially for medical image segmentation, this transfer solves existing explainability disadvantages. The code is available at &lt;a href='https://github.com/TillmannRheude/SegHiResGrad_CAM'&gt;https://github.com/TillmannRheude/SegHiResGrad_CAM&lt;/a&gt;

Weakly supervised semantic segmentation (WSSS) generally utilizes the Class Activation Map (CAM) to synthesize pseudo-labels. However, the current methods of obtaining CAM focus on salient features of a specific layer, resulting in highlighting the most discriminative regions and further leading to rough segmentation results for WSSS. In this paper, we rethink the potential of the ordinary classifier and find that if features of all the layers are applied, the classifier will obtain CAM with complete discriminative regions. Inspired by this, we propose Fully-CAM for WSSS, which can fully exploit the potential of the ordinary classifier and yield more accurate segmentation results. Precisely, Fully-CAM firstly weights feature with their corresponding gradients to yield CAMs of each layer, then fusing these layers’ CAMs could generate an ultimate CAM with complete discriminative regions. Furthermore, Fully-CAM is encapsulated into a plug-in, which can be mounted on any trained ordinary classifier with convolution layer, and it exceeds its previous performance without extra training.