Knowledge Distillation, as a model compression technique, has received great attention. The knowledge of a well-performed teacher is distilled to a student with a small architecture. The architecture of the small student is often chosen to be similar to their teacher's, with fewer layers or fewer channels, or both. However, even with the same number of FLOPs or parameters, the students with different architecture can achieve different generalization ability. The configuration of a student architecture requires intensive network architecture engineering. In this work, instead of designing a good student architecture manually, we propose to search for the optimal student automatically. Based on L1-norm optimization, a subgraph from the teacher network topology graph is selected as a student, the goal of which is to minimize the KL-divergence between student's and teacher's outputs. We verify the proposal on CIFAR10 and CIFAR100 datasets. The empirical experiments show that the learned student architecture achieves better performance than ones specified manually. We also visualize and understand the architecture of the found student.
New York University（纽约大学）；Shanghai Jiao Tong University（上海交通大学）；Rutgers University（罗格斯大学）
To promote better performance-bandwidth trade-off for multi-agent perception, we propose a novel distilled collaboration graph (DiscoGraph) to model trainable, pose-aware, and adaptive collaboration among agents. Our key novelties lie in two aspects. First, we propose a teacher-student framework to train DiscoGraph via knowledge distillation. The teacher model employs an early collaboration with holistic-view inputs; the student model is based on intermediate collaboration with single-view inputs. Our framework trains DiscoGraph by constraining post-collaboration feature maps in the student mo…
Simon Fraser University（西蒙弗雷泽大学）；Borealis AI（Borealis人工智能研究院）
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…
Lightweight Intrusion Detection in IoT via SHAP-Guided Feature Pruning and Knowledge-Distilled Kronecker Networks
LSIA Laboratory, Moroccan School of Engineering Sciences (EMSI)（摩洛哥工程科学学校LSIA实验室）；Laboratory of Research in Informatics (LaRI), Faculty of Sciences, Ibn Tofail University（伊本·泰菲勒大学信息研究实验室）；College of Computing, University Mohammed VI Polytechnic（穆莱·伊沙克六世理工学院计算机学院）
Videos shot by laymen using hand-held cameras contain undesirable shaky motion. Estimating the global motion between successive frames, in a manner not influenced by moving objects, is central to many video stabilization techniques, but poses significant challenges. A large body of work uses 2D affine transformations or homography for the global motion. However, in this work, we introduce a more general representation scheme, which adapts any existing optical flow network to ignore the moving objects and obtain a spatially smooth approximation of the global motion between video frames. We achi…
Feature Fusion and Knowledge-Distilled Multi-Modal Multi-Target Detection
School of Information and Communications, Hanoi University of Science and Technology（信息与通信学院，河内科学技术大学）；Telecom Neural Detection Lab, Polytechnique Montréal（电信神经检测实验室，蒙特利尔理工学院）
Since visual perception can give rich information beyond text descriptions for world understanding, there has been increasing interest in leveraging visual grounding for language learning. Recently, vokenization (Tan and Bansal, 2020) has attracted attention by using the predictions of a text-to-image retrieval model as labels for language model supervision. Despite its success, the method suffers from approximation error of using finite image labels and the lack of vocabulary diversity of a small image-text dataset. To overcome these limitations, we present VidLanKD, a video-language knowledg…
Towards Undistillable Models by Minimizing Conditional Mutual Information
Edward S. Rogers Sr. Department of Electrical and Computer Engineering University of Toronto（多伦多大学电气与计算机工程系）；Department of Electrical Engineering Stanford University（斯坦福大学电气工程系）；Department of Electrical and Computer Engineering University of Waterloo（滑铁卢大学电气与计算机工程系）
Hashing has been widely used in approximate nearest neighbor search for its storage and computational efficiency. Deep supervised hashing methods are not widely used because of the lack of labeled data, especially when the domain is transferred. Meanwhile, unsupervised deep hashing models can hardly achieve satisfactory performance due to the lack of reliable similarity signals. To tackle this problem, we propose a novel deep unsupervised hashing method, namely Distilled Smooth Guidance (DSG), which can learn a distilled dataset consisting of similarity signals as well as smooth confidence sig…
Lightweight Contrastive Distilled Hashing for Online Cross-modal Retrieval
School of Artificial Intelligence, Guangzhou University（广州大学人工智能学院）；School of Computer Science, Guangdong Polytechnic Normal University（广东技术师范大学计算机科学学院）；School of Automation, Guangdong University of Technology（广东工业大学自动化学院）；Bio-Computing Research Center, Harbin Institute of Technology（哈尔滨工业大学生物计算研究中心）
Today, transformer language models serve as a core component for majority of natural language processing tasks. Industrial application of such models requires minimization of computation time and memory footprint. Knowledge distillation is one of approaches to address this goal. Existing methods in this field are mainly focused on reducing the number of layers or dimension of embeddings/hidden representations. Alternative option is to reduce the number of tokens in vocabulary and therefore the embeddings matrix of the student model. The main problem with vocabulary minimization is mismatch bet…
Knowledge distillation is effective for producing small, high-performance neural networks for classification, but these small networks are vulnerable to adversarial attacks. This paper studies how adversarial robustness transfers from teacher to student during knowledge distillation. We find that a large amount of robustness may be inherited by the student even when distilled on only clean images. Second, we introduce Adversarially Robust Distillation (ARD) for distilling robustness onto student networks. In addition to producing small models with high test accuracy like conventional distillat…
University of Liverpool（利物浦大学）；Huawei Ireland Research Center（华为爱尔兰研究中心）
Pruning aims to reduce the number of parameters while maintaining performance close to the original network. This work proposes a novel \emph{self-distillation} based pruning strategy, whereby the representational similarity between the pruned and unpruned versions of the same network is maximized. Unlike previous approaches that treat distillation and pruning separately, we use distillation to inform the pruning criteria, without requiring a separate student network as in knowledge distillation. We show that the proposed {\em cross-correlation objective for self-distilled pruning} implicitly …
Novustech Services（诺沃斯特科技服务公司）；University of Bucharest（布加勒斯特大学）
In this paper, we study the task of facial expression recognition under strong occlusion. We are particularly interested in cases where 50% of the face is occluded, e.g. when the subject wears a Virtual Reality (VR) headset. While previous studies show that pre-training convolutional neural networks (CNNs) on fully-visible (non-occluded) faces improves the accuracy, we propose to employ knowledge distillation to achieve further improvements. First of all, we employ the classic teacher-student training strategy, in which the teacher is a CNN trained on fully-visible faces and the student is a C…
In Good GRACEs: Principled Teacher Selection for Knowledge Distillation
Princeton Language and Intelligence（普林斯顿语言与智能研究所）；Kempner Institute, Harvard（哈佛凯普纳研究所）；Microsoft Research, New York（微软研究院（纽约））；University of Pennsylvania（宾夕法尼亚大学）
Indian Institute of Technology, Jodhpur（焦特布尔印度理工学院）
Knowledge distillation is a common technique for improving the performance of a shallow student network by transferring information from a teacher network, which in general, is comparatively large and deep. These teacher networks are pre-trained and often uncalibrated, as no calibration technique is applied to the teacher model while training. Calibration of a network measures the probability of correctness for any of its predictions, which is critical in high-risk domains. In this paper, we study how to obtain a calibrated student from an uncalibrated teacher. Our approach relies on the fusio…
Knowledge distillation (KD) has gained much attention due to its effectiveness in compressing large-scale pre-trained models. In typical KD methods, the small student model is trained to match the soft targets generated by the big teacher model. However, the interaction between student and teacher is one-way. The teacher is usually fixed once trained, resulting in static soft targets to be distilled. This one-way interaction leads to the teacher's inability to perceive the characteristics of the student and its training progress. To address this issue, we propose Interactive Knowledge Distilla…
Distillation Traps and Guards: A Calibration Knob for LLM Distillability
Generative AI Lab, College of Computing and Data Science（生成人工智能实验室，计算与数据科学学院）；Nanyang Technological University（南洋理工大学）；Systems Research Institute of the Polish Academy of Sciences（波兰科学院系统研究所）；AGH University of Krakow（克拉科夫AGH大学）；SAN University（SAN大学）
Physics-Distilled Neural Network enabled by Large Language Models for Manufacturing Process-Property Predictive Modeling
School of Mechanical, Aerospace and Manufacturing Engineering, University of Connecticut（康奈尔大学机械、航空航天与制造工程学院）；Department of Mechanical & Aerospace Engineering, Rutgers, the State University of New Jersey（新泽西州立大学鲁特大学机械与航空航天工程学院）
This paper presents a novel knowledge distillation based model compression framework consisting of a student ensemble. It enables distillation of simultaneously learnt ensemble knowledge onto each of the compressed student models. Each model learns unique representations from the data distribution due to its distinct architecture. This helps the ensemble generalize better by combining every model's knowledge. The distilled students and ensemble teacher are trained simultaneously without requiring any pretrained weights. Moreover, our proposed method can deliver multi-compressed students with s…
National University of Defense Technology（国防科技大学）；Chinese Academy of Sciences（中国科学院）
Knowledge distillation (KD) is an effective training strategy to improve the lightweight student models under the guidance of cumbersome teachers. However, the large architecture difference across the teacher-student pairs limits the distillation gains. In contrast to previous adaptive distillation methods to reduce the teacher-student gap, we explore a novel training-free framework to search for the best student architectures for a given teacher. Our work first empirically show that the optimal model under vanilla training cannot be the winner in distillation. Secondly, we find that the simil…
Model distillation is frequently proposed as a technique to reduce the privacy leakage of machine learning. These empirical privacy defenses rely on the intuition that distilled ``student'' models protect the privacy of training data, as they only interact with this data indirectly through a ``teacher'' model. In this work, we design membership inference attacks to systematically study the privacy provided by knowledge distillation to both the teacher and student training sets. Our new attacks show that distillation alone provides only limited privacy across a number of domains. We explain the…
Xidian University（西安电子科技大学）；University of Sydney（悉尼大学）；Tencent AI Lab（腾讯人工智能实验室）
Due to the high storage and search efficiency, hashing has become prevalent for large-scale similarity search. Particularly, deep hashing methods have greatly improved the search performance under supervised scenarios. In contrast, unsupervised deep hashing models can hardly achieve satisfactory performance due to the lack of reliable supervisory similarity signals. To address this issue, we propose a novel deep unsupervised hashing model, dubbed DistillHash, which can learn a distilled data set consisted of data pairs, which have confidence similarity signals. Specifically, we investigate the…
TaxDistill: Improving Metagenomic Taxonomic Annotation via Distilled Genomic Foundation Models
National Genomics Data Center, China National Center for Bioinformation（中国生物信息中心国家基因组数据中心）；Beijing Key Laboratory of Intelligent Governance and Application of Biological Big Data, China National Center for Bioinformation（北京生物大数据智能治理与应用重点实验室，中国生物信息中心）；Beijing Institute of Genomics, Chinese Academy of Sciences（北京基因组研究所，中国科学院）；University of Chinese Academy of Sciences（中国科学院大学）
We propose a method to infer a dense depth map from a single image, its calibration, and the associated sparse point cloud. In order to leverage existing models (teachers) that produce putative depth maps, we propose an adaptive knowledge distillation approach that yields a positive congruent training process, wherein a student model avoids learning the error modes of the teachers. In the absence of ground truth for model selection and training, our method, termed Monitored Distillation, allows a student to exploit a blind ensemble of teachers by selectively learning from predictions that best…
Stanford University（斯坦福大学）；University of California, San Diego（加州大学圣地亚哥分校）
We present Knowledge Distillation with Meta Learning (MetaDistil), a simple yet effective alternative to traditional knowledge distillation (KD) methods where the teacher model is fixed during training. We show the teacher network can learn to better transfer knowledge to the student network (i.e., learning to teach) with the feedback from the performance of the distilled student network in a meta learning framework. Moreover, we introduce a pilot update mechanism to improve the alignment between the inner-learner and meta-learner in meta learning algorithms that focus on an improved inner-lea…
USC Information Sciences Institute（南加州大学信息科学研究所）；Google Research（谷歌研究院）
Despite the popularity and efficacy of knowledge distillation, there is limited understanding of why it helps. In order to study the generalization behavior of a distilled student, we propose a new theoretical framework that leverages supervision complexity: a measure of alignment between teacher-provided supervision and the student's neural tangent kernel. The framework highlights a delicate interplay among the teacher's accuracy, the student's margin with respect to the teacher predictions, and the complexity of the teacher predictions. Specifically, it provides a rigorous justification for …
Knowledge distillation is the procedure of transferring "knowledge" from a large model (the teacher) to a more compact one (the student), often being used in the context of model compression. When both models have the same architecture, this procedure is called self-distillation. Several works have anecdotally shown that a self-distilled student can outperform the teacher on held-out data. In this work, we systematically study self-distillation in a number of settings. We first show that even with a highly accurate teacher, self-distillation allows a student to surpass the teacher in all cases…
With the success of deep neural networks, knowledge distillation which guides the learning of a small student network from a large teacher network is being actively studied for model compression and transfer learning. However, few studies have been performed to resolve the poor learning issue of the student network when the student and teacher model sizes significantly differ. In this paper, we propose a densely guided knowledge distillation using multiple teacher assistants that gradually decreases the model size to efficiently bridge the large gap between the teacher and student networks. To…
The smaller memory bandwidth in smart devices prompts development of smaller Automatic Speech Recognition (ASR) models. To obtain a smaller model, one can employ the model compression techniques. Knowledge distillation (KD) is a popular model compression approach that has shown to achieve smaller model size with relatively lesser degradation in the model performance. In this approach, knowledge is distilled from a trained large size teacher model to a smaller size student model. Also, the transducer based models have recently shown to perform well for on-device streaming ASR task, while the co…
The University of Sydney（悉尼大学）；SenseTime Group Limited（商汤集团有限公司）
Knowledge distillation aims to transfer representation ability from a teacher model to a student model. Previous approaches focus on either individual representation distillation or inter-sample similarity preservation. While we argue that the inter-sample relation conveys abundant information and needs to be distilled in a more effective way. In this paper, we propose a novel knowledge distillation method, namely Complementary Relation Contrastive Distillation (CRCD), to transfer the structural knowledge from the teacher to the student. Specifically, we estimate the mutual relation in an anch…
Inception Institute of Artificial Intelligence（起源人工智能研究院）；Beijing Institute of Technology（北京理工大学）；Shandong University（山东大学）；Singapore Management University（新加坡管理大学）；Salesforce Research Asia（赛富时亚洲研究院）
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 st…
Knowledge distillation~(KD) is an effective learning paradigm for improving the performance of lightweight student networks by utilizing additional supervision knowledge distilled from teacher networks. Most pioneering studies either learn from only a single teacher in their distillation learning methods, neglecting the potential that a student can learn from multiple teachers simultaneously, or simply treat each teacher to be equally important, unable to reveal the different importance of teachers for specific examples. To bridge this gap, we propose a novel adaptive multi-teacher multi-level…
In natural language processing (NLP) tasks, slow inference speed and huge footprints in GPU usage remain the bottleneck of applying pre-trained deep models in production. As a popular method for model compression, knowledge distillation transfers knowledge from one or multiple large (teacher) models to a small (student) model. When multiple teacher models are available in distillation, the state-of-the-art methods assign a fixed weight to a teacher model in the whole distillation. Furthermore, most of the existing methods allocate an equal weight to every teacher model. In this paper, we obser…
Department of Electrical Engineering, POSTECH（POSTECH电气工程系）；Graduate School of Artificial Intelligence, POSTECH（POSTECH人工智能研究生院）；POSTECH（POSTECH（浦项科技大学））
Deep models are susceptible to learning spurious correlations, even during the post-processing. We take a closer look at the knowledge distillation -- a popular post-processing technique for model compression -- and find that distilling with biased training data gives rise to a biased student, even when the teacher is debiased. To address this issue, we propose a simple knowledge distillation algorithm, coined DeTT (Debiasing by Teacher Transplanting). Inspired by a recent observation that the last neural net layer plays an overwhelmingly important role in debiasing, DeTT directly transplants …
Pre-trained language models achieve outstanding performance in NLP tasks. Various knowledge distillation methods have been proposed to reduce the heavy computation and storage requirements of pre-trained language models. However, from our observations, student models acquired by knowledge distillation suffer from adversarial attacks, which limits their usage in security sensitive scenarios. In order to overcome these security problems, RoSearch is proposed as a comprehensive framework to search the student models with better adversarial robustness when performing knowledge distillation. A dire…
The University of Queensland（昆士兰大学）；CSIRO Data61（澳大利亚联邦科学与工业研究组织Data61部门）
In knowledge distillation, previous feature distillation methods mainly focus on the design of loss functions and the selection of the distilled layers, while the effect of the feature projector between the student and the teacher remains under-explored. In this paper, we first discuss a plausible mechanism of the projector with empirical evidence and then propose a new feature distillation method based on a projector ensemble for further performance improvement. We observe that the student network benefits from a projector even if the feature dimensions of the student and the teacher are the …
