Knowledge distillation is a simple but powerful way to transfer knowledge between a teacher model to a student model. Existing work suffers from at least one of the following key limitations in terms of direction and scope of transfer which restrict its use: all knowledge is transferred from teacher to student regardless of whether or not that knowledge is useful, the student is the only one learning in this exchange, and typically distillation transfers knowledge only from a single teacher to a single student. We formulate a novel form of knowledge distillation in which many models can act as both students and teachers which we call cooperative distillation. The models cooperate as follows: a model (the student) identifies specific deficiencies in it's performance and searches for another model (the teacher) who encodes learned knowledge into instructional virtual instances via counterfactual instance generation. Because different models may have different strengths and weaknesses, all models can act as either students or teachers (cooperation) when appropriate and only distill knowledge in areas specific to their strengths (focus). Since counterfactuals as a paradigm are not tied to any specific algorithm, we can use this method to distill knowledge between learners of different architectures, algorithms, and even feature spaces. We demonstrate our approach not only outperforms baselines such as transfer learning, self-supervised learning, and multiple knowledge distillation algorithms on several datasets, but it can also be used in settings where the aforementioned techniques cannot.
论文检索
输入标题、作者或关键词，从 2,578 篇学术成果中精准定位
Semi-supervised time-series classification could effectively alleviate the issue of lacking labeled data. However, existing approaches usually ignore model interpretability, making it difficult for humans to understand the principles behind the predictions of a model. Shapelets are a set of discriminative subsequences that show high interpretability in time series classification tasks. Shapelet learning-based methods have demonstrated promising classification performance. Unfortunately, without enough labeled data, the shapelets learned by existing methods are often poorly discriminative, and even dissimilar to any subsequence of the original time series. To address this issue, we propose the Diffusion Language-Shapelets model (DiffShape) for semi-supervised time series classification. In DiffShape, a self-supervised diffusion learning mechanism is designed, which uses real subsequences as a condition. This helps to increase the similarity between the learned shapelets and real subsequences by using a large amount of unlabeled data. Furthermore, we introduce a contrastive language-shapelets learning strategy that improves the discriminability of the learned shapelets by incorporating the natural language descriptions of the time series. Experiments have been conducted on the UCR time series archive, and the results reveal that the proposed DiffShape method achieves state-of-the-art performance and exhibits superior interpretability over baselines.
Clustering methods achieve performance improvement by jointly learning representation and cluster assignment. However, they do not consider the confidence of pseudo-labels which are not optimal as supervised information, resulting into error accumulation. To address this issue, we propose a Robust Pseudo-labeling for Semantic Clustering (RPSC) approach, which includes two stages. In the first stage (RPSC-Self), we design a semantic pseudo-labeling scheme by using the consistency of samples, i.e., samples with same semantics should be close to each other in the embedding space. To exploit robust semantic pseudo-labels for self-supervised learning, we propose a soft contrastive loss (SCL) which encourage the model to believe high-confidence sematic pseudo-labels and be less driven by low-confidence pseudo-labels. In the second stage (RPSC-Semi), we first determine the semantic pseudo-label of a sample based on the distance between itself and cluster centers, followed by screening out reliable semantic pseudo-label by exploiting the consistency. These reliable pseudo-labels are used as supervised information in the pseudo-semi-supervised learning algorithm to further improve the performance. Experimental results show that RPSC outperforms 18 competitive clustering algorithms significantly on six challenging image benchmarks. In particular, RPSC achieves an accuracy of 0.688 on ImageNet-Dogs, which is an up to 24% improvement, compared with the second-best method. Meanwhile, we conduct ablation studies to investigate effects of different augmented strategies on RPSC as well as contributions of terms in SCL to clustering performance. Besides, experimental results indicate that SCL can be easily integrated into existing clustering methods and bring performance improvement.
Graph Comparative Learning (GCL) is a self-supervised method that combines the advantages of Graph Convolutional Networks (GCNs) and comparative learning, making it promising for learning node representations. However, the GCN encoders used in these methods rely on the Fourier transform to learn fixed graph representations, which is inherently limited by the uncertainty principle involving spatial and spectral localization trade-offs. To overcome the inflexibility of existing methods and the computationally expensive eigen-decomposition and dense matrix multiplication, this paper proposes an Adaptive Spectral Wavelet Transform-based Self-Supervised Graph Neural Network (ASWT-SGNN). The proposed method employs spectral adaptive polynomials to approximate the filter function and optimize the wavelet using contrast loss. This design enables the creation of local filters in both spectral and spatial domains, allowing flexible aggregation of neighborhood information at various scales and facilitating controlled transformation between local and global information. Compared to existing methods, the proposed approach reduces computational complexity and addresses the limitation of graph convolutional neural networks, which are constrained by graph size and lack flexible control over the neighborhood aspect. Extensive experiments on eight benchmark datasets demonstrate that ASWT-SGNN accurately approximates the filter function in high-density spectral regions, avoiding costly eigen-decomposition. Furthermore, ASWT-SGNN achieves comparable performance to state-of-the-art models in node classification tasks.
TimesURL: Self-Supervised Contrastive Learning for Universal Time Series Representation Learning
PDF ↗Learning universal time series representations applicable to various types of downstream tasks is challenging but valuable in real applications. Recently, researchers have attempted to leverage the success of self-supervised contrastive learning (SSCL) in Computer Vision(CV) and Natural Language Processing(NLP) to tackle time series representation. Nevertheless, due to the special temporal characteristics, relying solely on empirical guidance from other domains may be ineffective for time series and difficult to adapt to multiple downstream tasks. To this end, we review three parts involved in SSCL including 1) designing augmentation methods for positive pairs, 2) constructing (hard) negative pairs, and 3) designing SSCL loss. For 1) and 2), we find that unsuitable positive and negative pair construction may introduce inappropriate inductive biases, which neither preserve temporal properties nor provide sufficient discriminative features. For 3), just exploring segment- or instance-level semantics information is not enough for learning universal representation. To remedy the above issues, we propose a novel self-supervised framework named TimesURL. Specifically, we first introduce a frequency-temporal-based augmentation to keep the temporal property unchanged. And then, we construct double Universums as a special kind of hard negative to guide better contrastive learning. Additionally, we introduce time reconstruction as a joint optimization objective with contrastive learning to capture both segment-level and instance-level information. As a result, TimesURL can learn high-quality universal representations and achieve state-of-the-art performance in 6 different downstream tasks, including short- and long-term forecasting, imputation, classification, anomaly detection and transfer learning.
Deep cross-modal hashing technology provides an effective and efficient cross-modal unified representation learning solution for cross-modal search. However, the existing methods neglect the implicit fine-grained multimodal knowledge relations between these modalities such as when the image contains information that is not directly described in the text. To tackle this problem, we propose a novel self-supervised multi-grained multi-modal knowledge graph contrastive hashing method for cross-modal search (CMGCH). Firstly, in order to capture implicit fine-grained cross-modal semantic associations, a multi-modal knowledge graph is constructed, which represents the implicit multimodal knowledge relations between the image and text as inter-modal and intra-modal semantic associations. Secondly, a cross-modal graph contrastive attention network is proposed to reason on the multi-modal knowledge graph to sufficiently learn the implicit fine-grained inter-modal and intra-modal knowledge relations. Thirdly, a cross-modal multi-granularity contrastive embedding learning mechanism is proposed, which fuses the global coarse-grained and local fine-grained embeddings by multihead attention mechanism for inter-modal and intra-modal contrastive learning, so as to enhance the cross-modal unified representations with stronger discriminativeness and semantic consistency preserving power. With the joint training of intra-modal and inter-modal contrast, the invariant and modal-specific information of different modalities can be maintained in the final unified cross-modal unified hash space. Extensive experiments on several cross-modal benchmark datasets demonstrate that the proposed CMGCH outperforms the state-of the-art methods.
When Model Meets New Normals: Test-Time Adaptation for Unsupervised Time-Series Anomaly Detection
PDF ↗Time-series anomaly detection deals with the problem of detecting anomalous timesteps by learning normality from the sequence of observations. However, the concept of normality evolves over time, leading to a "new normal problem", where the distribution of normality can be changed due to the distribution shifts between training and test data. This paper highlights the prevalence of the new normal problem in unsupervised time-series anomaly detection studies. To tackle this issue, we propose a simple yet effective test-time adaptation strategy based on trend estimation and a self-supervised approach to learning new normalities during inference. Extensive experiments on real-world benchmarks demonstrate that incorporating the proposed strategy into the anomaly detector consistently improves the model's performances compared to the existing baselines, leading to robustness to the distribution shifts.
Measuring Self-Supervised Representation Quality for Downstream Classification Using Discriminative Features
PDF ↗Self-supervised learning (SSL) has shown impressive results in downstream classification tasks. However, there is limited work in understanding their failure modes and interpreting their learned representations. In this paper, we study the representation space of state-of-the-art self-supervised models including SimCLR, SwaV, MoCo, BYOL, DINO, SimSiam, VICReg and Barlow Twins. Without the use of class label information, we discover discriminative features that correspond to unique physical attributes in images, present mostly in correctly-classified representations. Using these features, we can compress the representation space by up to$40% without significantly affecting linear classification performance. We then propose Self-Supervised Representation Quality Score (or Q-Score), an unsupervised score that can reliably predict if a given sample is likely to be mis-classified during linear evaluation, achieving AUPRC of 91.45 on ImageNet-100 and 78.78 on ImageNet-1K. Q-Score can also be used as a regularization term on pre-trained encoders to remedy low-quality representations. Fine-tuning with Q-Score regularization can boost the linear probing accuracy of SSL models by up to 5.8% on ImageNet-100 and 3.7% on ImageNet-1K compared to their baselines. Finally, using gradient heatmaps and Salient ImageNet masks, we define a metric to quantify the interpretability of each representation. We show that discriminative features are strongly correlated to core attributes and, enhancing these features through Q-score regularization makes SSL representations more interpretable.
A fundamental challenge of bipartite graph representation learning is how to extract informative node embeddings. Self-Supervised Learning (SSL) is a promising paradigm to address this challenge. Most recent bipartite graph SSL methods are based on contrastive learning which learns embeddings by discriminating positive and negative node pairs. Contrastive learning usually requires a large number of negative node pairs, which could lead to computational burden and semantic errors. In this paper, we introduce a novel synergistic representation learning model (STERLING) to learn node embeddings without negative node pairs. STERLING preserves the unique local and global synergies in bipartite graphs. The local synergies are captured by maximizing the similarity of the inter-type and intra-type positive node pairs, and the global synergies are captured by maximizing the mutual information of co-clusters. Theoretical analysis demonstrates that STERLING could improve the connectivity between different node types in the embedding space. Extensive empirical evaluation on various benchmark datasets and tasks demonstrates the effectiveness of STERLING for extracting node embeddings.
Recent works attempt to extend Graph Convolution Networks (GCNs) to point clouds for classification and segmentation tasks. These works tend to sample and group points to create smaller point sets locally and mainly focus on extracting local features through GCNs, while ignoring the relationship between point sets. In this paper, we propose the Dynamic Hop Graph Convolution Network (DHGCN) for explicitly learning the contextual relationships between the voxelized point parts, which are treated as graph nodes. Motivated by the intuition that the contextual information between point parts lies in the pairwise adjacent relationship, which can be depicted by the hop distance of the graph quantitatively, we devise a novel self-supervised part-level hop distance reconstruction task and design a novel loss function accordingly to facilitate training. In addition, we propose the Hop Graph Attention (HGA), which takes the learned hop distance as input for producing attention weights to allow edge features to contribute distinctively in aggregation. Eventually, the proposed DHGCN is a plug-and-play module that is compatible with point-based backbone networks. Comprehensive experiments on different backbones and tasks demonstrate that our self-supervised method achieves state-of-the-art performance. Our source codes are available at: https://github.com/Jinec98/DHGCN.
Video-based facial analysis is important for autonomous agents to understand human expressions and sentiments. However, limited labeled data is available to learn effective facial representations. This paper proposes a novel self-supervised face-centric pretraining framework, called PrefAce, which learns transferable video facial representation without labels. The self-supervised learning is performed with an effective landmark-guided global-local tube distillation. Meanwhile, a novel instance-wise update FaceFeat Cache is built to enforce more discriminative and diverse representations for downstream tasks. Extensive experiments demonstrate that the proposed framework learns universal instance-aware facial representations with fine-grained landmark details from videos. The point is that it can transfer across various facial analysis tasks, e.g., Facial Attribute Recognition (FAR), Facial Expression Recognition (FER), DeepFake Detection (DFD), and Lip Synchronization (LS). Our framework also outperforms the state-of-the-art on various downstream tasks, even in low data regimes. Code is available at https://github.com/siyuan-h/PrefAce.
Transitivity-Preserving Graph Representation Learning for Bridging Local Connectivity and Role-Based Similarity
PDF ↗Graph representation learning (GRL) methods, such as graph neural networks and graph transformer models, have been successfully used to analyze graph-structured data, mainly focusing on node classification and link prediction tasks. However, the existing studies mostly only consider local connectivity while ignoring long-range connectivity and the roles of nodes. In this paper, we propose Unified Graph Transformer Networks (UGT) that effectively integrate local and global structural information into fixed-length vector representations. First, UGT learns local structure by identifying the local sub-structures and aggregating features of the k-hop neighborhoods of each node. Second, we construct virtual edges, bridging distant nodes with structural similarity to capture the long-range dependencies. Third, UGT learns unified representations through self-attention, encoding structural distance and p-step transition probability between node pairs. Furthermore, we propose a self-supervised learning task that effectively learns transition probability to fuse local and global structural features, which could then be transferred to other downstream tasks. Experimental results on real-world benchmark datasets over various downstream tasks showed that UGT significantly outperformed baselines that consist of state-of-the-art models. In addition, UGT reaches the third-order Weisfeiler-Lehman power to distinguish non-isomorphic graph pairs.
Graph contrastive learning (GCL) has attracted considerable attention because it can self-supervisedly extract low-dimensional representation of graph data. InfoNCE-based loss function is widely used in graph contrastive learning, which pulls the representations of positive pairs close to each other and pulls the representations of negative pairs away from each other. Recent works mainly focus on designing new augmentation methods or sampling strategies. However, we argue that the widely used InfoNCE-based methods may contain an implicit conflict which seriously confuses models when learning from negative pairs. This conflict is engendered by the encoder's message-passing mechanism and the InfoNCE loss function. As a result, the learned representations between negative samples cannot be far away from each other, compromising the model performance. To our best knowledge, this is the first time to report and analysis this conflict of GCL. To address this problem, we propose a simple but effective method called Partial ignored Graph Contrastive Learning (PiGCL). Specifically, PiGCL first dynamically captures the conflicts during training by detecting the gradient of representation similarities. It then enables the loss function to ignore the conflict, allowing the encoder to adaptively learn the ignored information without self-supervised samples. Extensive experiments demonstrate the effectiveness of our method.
Partial multi-view clustering is a challenging and practical research problem for data analysis in real-world applications, due to the potential data missing issue in different views. However, most existing methods have not fully explored the correlation information among various incomplete views. In addition, these existing clustering methods always ignore discovering discriminative features inside the data itself in this unsupervised task. To tackle these challenges, we propose Partial Multi-View Clustering via Self-Supervised \textbf{N}etwork (PVC-SSN) in this paper. Specifically, we employ contrastive learning to obtain a more discriminative and consistent subspace representation, which is guided by a self-supervised module. Self-supervised learning can exploit effective cluster information through the data itself to guide the learning process of clustering tasks. Thus, it can pull together embedding features from the same cluster and push apart these from different clusters. Extensive experiments on several benchmark datasets show that the proposed PVC-SCN method outperforms several state-of-the-art clustering methods.
Unsupervised and self-supervised representation learning has become popular in recent years for learning useful features from unlabelled data. Representation learning has been mostly developed in the neural network literature, and other models for representation learning are surprisingly unexplored. In this work, we introduce and analyze several kernel-based representation learning approaches: Firstly, we define two kernel Self-Supervised Learning (SSL) models using contrastive loss functions and secondly, a Kernel Autoencoder (AE) model based on the idea of embedding and reconstructing data. We argue that the classical representer theorems for supervised kernel machines are not always applicable for (self-supervised) representation learning, and present new representer theorems, which show that the representations learned by our kernel models can be expressed in terms of kernel matrices. We further derive generalisation error bounds for representation learning with kernel SSL and AE, and empirically evaluate the performance of these methods in both small data regimes as well as in comparison with neural network based models.
Recent machine learning algorithms have been developed using well-curated datasets, which often require substantial cost and resources. On the other hand, the direct use of raw data often leads to overfitting towards frequently occurring class information. To address class imbalances cost-efficiently, we propose an active data filtering process during self-supervised pre-training in our novel framework, Duplicate Elimination (DUEL). This framework integrates an active memory inspired by human working memory and introduces distinctiveness information, which measures the diversity of the data in the memory, to optimize both the feature extractor and the memory. The DUEL policy, which replaces the most duplicated data with new samples, aims to enhance the distinctiveness information in the memory and thereby mitigate class imbalances. We validate the effectiveness of the DUEL framework in class-imbalanced environments, demonstrating its robustness and providing reliable results in downstream tasks. We also analyze the role of the DUEL policy in the training process through various metrics and visualizations.
Fed-QSSL: A Framework for Personalized Federated Learning under Bitwidth and Data Heterogeneity
PDF ↗Motivated by high resource costs of centralized machine learning schemes as well as data privacy concerns, federated learning (FL) emerged as an efficient alternative that relies on aggregating locally trained models rather than collecting clients' potentially private data. In practice, available resources and data distributions vary from one client to another, creating an inherent system heterogeneity that leads to deterioration of the performance of conventional FL algorithms. In this work, we present a federated quantization-based self-supervised learning scheme (Fed-QSSL) designed to address heterogeneity in FL systems. At clients' side, to tackle data heterogeneity we leverage distributed self-supervised learning while utilizing low-bit quantization to satisfy constraints imposed by local infrastructure and limited communication resources. At server's side, Fed-QSSL deploys de-quantization, weighted aggregation and re-quantization, ultimately creating models personalized to both data distribution as well as specific infrastructure of each client's device. We validated the proposed algorithm on real world datasets, demonstrating its efficacy, and theoretically analyzed impact of low-bit training on the convergence and robustness of the learned models.
Visual Reinforcement Learning (RL) is a promising approach to achieve human-like intelligence. However, it currently faces challenges in learning efficiently within noisy environments. In contrast, humans can quickly identify task-relevant objects in distraction-filled surroundings by applying previously acquired common knowledge. Recently, foundational models in natural language processing and computer vision have achieved remarkable successes, and the common knowledge within these models can significantly benefit downstream task training. Inspired by these achievements, we aim to incorporate common knowledge from foundational models into visual RL. We propose a novel Focus-Then-Decide (FTD) framework, allowing the agent to make decisions based solely on task-relevant objects. To achieve this, we introduce an attention mechanism to select task-relevant objects from the object set returned by a foundational segmentation model, and only use the task-relevant objects for the subsequent training of the decision module. Additionally, we specifically employed two generic self-supervised objectives to facilitate the rapid learning of this attention mechanism. Experimental results on challenging tasks based on DeepMind Control Suite and Franka Emika Robotics demonstrate that our method can quickly and accurately pinpoint objects of interest in noisy environments. Consequently, it achieves a significant performance improvement over current state-of-the-art algorithms. Project Page: https://www.lamda.nju.edu.cn/chenc/FTD.html Code: https://github.com/LAMDA-RL/FTD
Few-shot classification (FSC) entails learning novel classes given only a few examples per class after a pre-training (or meta-training) phase on a set of base classes. Recent works have shown that simply fine-tuning a pre-trained Vision Transformer (ViT) on new test classes is a strong approach for FSC. Fine-tuning ViTs, however, is expensive in time, compute and storage. This has motivated the design of parameter efficient fine-tuning (PEFT) methods which fine-tune only a fraction of the Transformer's parameters. While these methods have shown promise, inconsistencies in experimental conditions make it difficult to disentangle their advantage from other experimental factors including the feature extractor architecture, pre-trained initialization and fine-tuning algorithm, amongst others. In our paper, we conduct a large-scale, experimentally consistent, empirical analysis to study PEFTs for few-shot image classification. Through a battery of over 1.8k controlled experiments on large-scale few-shot benchmarks including Meta-Dataset and ORBIT, we uncover novel insights on PEFTs that cast light on their efficacy in fine-tuning ViTs for few-shot classification. Through our controlled empirical study, we have two main findings: (i) Fine-tuning just the LayerNorm parameters (which we call LN-Tune) during few-shot adaptation is an extremely strong baseline across ViTs pre-trained with both self-supervised and supervised objectives, (ii) For self-supervised ViTs, we find that simply learning a set of scaling parameters for each attention matrix (which we call Attn-Scale) along with a domain-residual adapter (DRA) module leads to state-of-the-art performance (while being ~9x more parameter-efficient) on Meta-Dataset. Our empirical findings set strong baselines and call for rethinking the current design of PEFT methods for FSC.
Probabilistic circuits (PCs) such as sum-product networks efficiently represent large multi-variate probability distributions. They are preferred in practice over other probabilistic representations, such as Bayesian and Markov networks, because PCs can solve marginal inference (MAR) tasks in time that scales linearly in the size of the network. Unfortunately, the most probable explanation (MPE) task and its generalization, the marginal maximum-a-posteriori (MMAP) inference task remain NP-hard in these models. Inspired by the recent work on using neural networks for generating near-optimal solutions to optimization problems such as integer linear programming, we propose an approach that uses neural networks to approximate MMAP inference in PCs. The key idea in our approach is to approximate the cost of an assignment to the query variables using a continuous multilinear function and then use the latter as a loss function. The two main benefits of our new method are that it is self-supervised, and after the neural network is learned, it requires only linear time to output a solution. We evaluate our new approach on several benchmark datasets and show that it outperforms three competing linear time approximations: max-product inference, max-marginal inference, and sequential estimation, which are used in practice to solve MMAP tasks in PCs.