Both long-tailed and noisily labeled data frequently appear in real-world applications and impose significant challenges for learning. Most prior works treat either problem in an isolated way and do not explicitly consider the coupling effects of the two. Our empirical observation reveals that such solutions fail to consistently improve the learning when the dataset is long-tailed with label noise. Moreover, with the presence of label noise, existing methods do not observe universal improvements across different sub-populations; in other words, some sub-populations enjoyed the benefits of improved accuracy at the cost of hurting others. Based on these observations, we introduce the Fairness Regularizer (FR), inspired by regularizing the performance gap between any two sub-populations. We show that the introduced fairness regularizer improves the performances of sub-populations on the tail and the overall learning performance. Extensive experiments demonstrate the effectiveness of the proposed solution when complemented with certain existing popular robust or class-balanced methods.

Pseudo-labeling has emerged as a popular and effective approach for utilizing unlabeled data. However, in the context of semi-supervised multi-label learning (SSMLL), conventional pseudo-labeling methods encounter difficulties when dealing with instances associated with multiple labels and an unknown label count. These limitations often result in the introduction of false positive labels or the neglect of true positive ones. To overcome these challenges, this paper proposes a novel solution called Class-distribution-Aware Pseudo-labeling (CAP) that performs pseudo-labeling in a class-aware manner. The proposed approach introduces a regularized learning framework incorporating class-aware thresholds, which effectively control the assignment of positive and negative pseudo-labels for each class. Notably, even with a small proportion of labeled examples, our observations demonstrate that the estimated class distribution serves as a reliable approximation. Motivated by this finding, we develop a class-distribution-aware thresholding (CAT) strategy to ensure the alignment of pseudo-label distribution with the true distribution. Moreover, we extend CAT into a label decision method, aiming to improve the model's classification performance during the testing phase. The correctness of the estimated class distribution is theoretically verified, and a generalization error bound is provided for our proposed method. Extensive experiments on multiple benchmark datasets confirm the efficacy of CAP in addressing the challenges of SSMLL problems.

Recently, Graph Transformer (GT) models have been widely used in the task of Molecular Property Prediction (MPP) due to their high reliability in characterizing the latent relationship among graph nodes (i.e., the atoms in a molecule). However, most existing GT-based methods usually explore the basic interactions between pairwise atoms, and thus they fail to consider the important interactions among critical motifs (e.g., functional groups consisted of several atoms) of molecules. As motifs in a molecule are significant patterns that are of great importance for determining molecular properties (e.g., toxicity and solubility), overlooking motif interactions inevitably hinders the effectiveness of MPP. To address this issue, we propose a novel Atom-Motif Contrastive Transformer (AMCT), which not only explores the atom-level interactions but also considers the motif-level interactions. Since the representations of atoms and motifs for a given molecule are actually two different views of the same instance, they are naturally aligned to generate the self-supervisory signals for model training. Meanwhile, the same motif can exist in different molecules, and hence we also employ the contrastive loss to maximize the representation agreement of identical motifs across different molecules. Finally, in order to clearly identify the motifs that are critical in deciding the properties of each molecule, we further construct a property-aware attention mechanism into our learning framework. Our proposed AMCT is extensively evaluated on 10 popular benchmark datasets, and both quantitative and qualitative results firmly demonstrate its effectiveness when compared with the state-of-the-art methods.

Recently, multi-view learning (MVL) has garnered significant attention due to its ability to fuse discriminative information from multiple views. However, real-world multi-view datasets are often heterogeneous and imperfect, which usually causes MVL methods designed for specific combinations of views to lack application potential and limits their effectiveness. To address this issue, we propose a novel robust MVL method (namely RML) with simultaneous representation fusion and alignment. Specifically, we introduce a simple yet effective multi-view transformer fusion network where we transform heterogeneous multi-view data into homogeneous word embeddings, and then integrate multiple views by the sample-level attention mechanism to obtain a fused representation. Furthermore, we propose a simulated perturbation based multi-view contrastive learning framework that dynamically generates the noise and unusable perturbations for simulating imperfect data conditions. The simulated noisy and unusable data obtain two distinct fused representations, and we utilize contrastive learning to align them for learning discriminative and robust representations. Our RML is self-supervised and can also be applied for downstream tasks as a regularization. In experiments, we employ it in multi-view unsupervised clustering, noise-label classification, and as a plug-and-play module for cross-modal hashing retrieval. Extensive comparison experiments and ablation studies validate RML's effectiveness. Code is available at https://github.com/SubmissionsIn/RML.

Partial-label learning (PLL) is a weakly supervised learning problem in which each example is associated with multiple candidate labels and only one is the true label. In recent years, many deep PLL algorithms have been developed to improve model performance. However, we find that some early developed algorithms are often underestimated and can outperform many later algorithms with complicated designs. In this paper, we delve into the empirical perspective of PLL and identify several critical but previously overlooked issues. First, model selection for PLL is non-trivial, but has never been systematically studied. Second, the experimental settings are highly inconsistent, making it difficult to evaluate the effectiveness of the algorithms. Third, there is a lack of real-world image datasets that can be compatible with modern network architectures. Based on these findings, we propose PLENCH, the first Partial-Label learning bENCHmark to systematically compare state-of-the-art deep PLL algorithms. We investigate the model selection problem for PLL for the first time, and propose novel model selection criteria with theoretical guarantees. We also create Partial-Label CIFAR-10 (PLCIFAR10), an image dataset of human-annotated partial labels collected from Amazon Mechanical Turk, to provide a testbed for evaluating the performance of PLL algorithms in more realistic scenarios. Researchers can quickly and conveniently perform a comprehensive and fair evaluation and verify the effectiveness of newly developed algorithms based on PLENCH. We hope that PLENCH will facilitate standardized, fair, and practical evaluation of PLL algorithms in the future.

The world is understood from various modalities, such as appearance, sound, language, etc. Since each modality only partially represents objects in a certain physical meaning, leveraging additional ones is beneficial in both theory and practice. However, exploiting novel modalities normally requires cross-modal pairs corresponding to the same instance, which is extremely resource-consuming and sometimes even impossible, making knowledge exploration of novel modalities largely restricted. To seek practical multi-modal learning, here we study Out-of-Modal (OOM) Generalization as an initial attempt to generalize to an unknown modality without given instance-level modal correspondence. Specifically, we consider Semi-Supervised and Unsupervised scenarios of OOM Generalization, where the first has scarce correspondences and the second has none, and propose connect & explore (COX) to solve these problems. COX first connects OOM data and known In-Modal (IM) data through a variational information bottleneck framework to extract shared information. Then, COX leverages the shared knowledge to create emergent correspondences, which is theoretically justified from an information-theoretic perspective. As a result, the label information on OOM data emerges along with the correspondences, which help explore the OOM data with unknown knowledge, thus benefiting generalization results. We carefully evaluate the proposed COX method under various OOM generalization scenarios, verifying its effectiveness and extensibility.

Robotic manipulation tasks often rely on visual inputs from cameras to perceive the environment. However, previous approaches still suffer from performance degradation when the camera’s viewpoint changes during manipulation. In this paper, we propose ReViWo (Representation learning for View-invariant World model), leveraging multi-view data to learn robust representations for control under viewpoint disturbance. ReViWo utilizes an autoencoder framework to reconstruct target images by an architecture that combines view-invariant representation (VIR) and view-dependent representation. To train ReViWo, we collect multi-view data in simulators with known view labels, meanwhile, ReViWo is simutaneously trained on Open X-Embodiment datasets without view labels. The VIR is then used to train a world model on pre-collected manipulation data and a policy through interaction with the world model. We evaluate the effectiveness of ReViWo in various viewpoint disturbance scenarios, including control under novel camera positions and frequent camera shaking, using the Meta-world & PandaGym environments. Besides, we also conduct experiments on real world ALOHA robot. The results demonstrate that ReViWo maintains robust performance under viewpoint disturbance, while baseline methods suffer from significant performance degradation. Furthermore, we show that the VIR captures task-relevant state information and remains stable for observations from novel viewpoints, validating the efficacy of the ReViWo approach.

Deep networks are prone to catastrophic forgetting during sequential task learning, i.e., losing the knowledge about old tasks upon learning new tasks. To this end, continual learning (CL) has emerged, whose existing methods focus mostly on regulating or protecting the parameters associated with the previous tasks. However, parameter protection is often impractical, since the size of parameters for storing the old-task knowledge increases linearly with the number of tasks, otherwise it is hard to preserve the parameters related to the old-task knowledge. In this work, we bring a dual opinion from neuroscience and physics to CL: in the whole networks, the pathways matter more than the parameters when concerning the knowledge acquired from the old tasks. Following this opinion, we propose a novel CL framework, learning without isolation (LwI), where model fusion is formulated as graph matching and the pathways occupied by the old tasks are protected without being isolated. Thanks to the sparsity of activation channels in a deep network, LwI can adaptively allocate available pathways for a new task, realizing pathway protection and addressing catastrophic forgetting in a parameter-effcient manner. Experiments on popular benchmark datasets demonstrate the superiority of the proposed LwI.

Deep learning with noisy labels presents significant challenges. In this work, we theoretically characterize the role of label noise from a feature learning perspective. Specifically, we consider a signal-noise data distribution, where each sample comprises a label-dependent signal and label-independent noise, and rigorously analyze the training dynamics of a two-layer convolutional neural network under this data setup, along with the presence of label noise. Our analysis identifies two key stages. In Stage I, the model perfectly fits all the clean samples (i.e., samples without label noise) while ignoring the noisy ones (i.e., samples with noisy labels). During this stage, the model learns the signal from the clean samples, which generalizes well on unseen data. In Stage II, as the training loss converges, the gradient in the direction of noise surpasses that of the signal, leading to overfitting on noisy samples. Eventually, the model memorizes the noise present in the noisy samples and degrades its generalization ability. Furthermore, our analysis provides a theoretical basis for two widely used techniques for tackling label noise: early stopping and sample selection. Experiments on both synthetic and real-world setups validate our theory.

With the growing deployment of large language models (LLMs) across diverse domains, concerns regarding their safety have grown substantially. LLM unlearning has emerged as a pivotal approach to removing harmful or unlawful contents while maintaining utility. Despite increasing interest, the challenges of continual unlearning, which is common in real-world scenarios, remain underexplored. Successive unlearning tasks often lead to intensified utility degradation. To effectively unlearn targeted knowledge while preserving LLM utility, it is essential to minimize changes in model parameters by selectively updating those linked to the target knowledge, thereby ensuring other knowledge remains unaffected. Building on the task vector framework, we propose a new method named ALKN (Adaptive Localization of Knowledge Negation), which uses dynamic masking to sparsify training gradients and adaptively adjusts unlearning intensity based on inter-task relationships. Comprehensive experiments across three well-established LLM unlearning datasets demonstrate that our approach consistently outperforms baseline methods in both unlearning effectiveness and utility retention under continual unlearning settings.

Real-world data may contain a considerable amount of noisily labeled examples, which usually mislead the training algorithm and result in degraded classification performance on test data. Therefore, Label Noise Learning (LNL) was proposed, of which one popular research trend focused on estimating the critical statistics (e.g., sample mean and sample covariance), to recover the clean data distribution. However, existing methods may suffer from the unreliable sample selection process or can hardly be applied to multi-class cases. Inspired by the centroid estimation theory, we propose Per-Class Statistic Estimation (PCSE), which establishes the quantitative relationship between the clean (first-order and second-order) statistics and the corresponding noisy statistics for every class. This relationship is further utilized to induce a generative classifier for model inference. Unlike existing methods, our approach does not require sample selection from the instance level. Moreover, our PCSE can serve as a general post-processing strategy applicable to various popular networks pre-trained on the noisy dataset for boosting their classification performance. Theoretically, we prove that the estimated statistics converge to their ground-truth values as the sample size increases, even if the estimated label transition matrix is biased. Empirically, we conducted intensive experiments on various binary and multi-class datasets, and the results demonstrate that PCSE achieves more precise statistic estimation as well as higher classification accuracy when compared with state-of-the-art methods in LNL.

Multi-view clustering (MVC) aims at exploring category structures among multi-view data in self-supervised manners. Multiple views provide more information than single views and thus existing MVC methods can achieve satisfactory performance. However, their performance might seriously degenerate when the views are noisy in practical multi-view scenarios. In this paper, we formally investigate the drawback of noisy views and then propose a theoretically grounded deep MVC method (namely MVCAN) to address this issue. Specifically, we propose a novel MVC objective that enables un-shared parameters and inconsistent clustering predictions across multiple views to reduce the side effects of noisy views. Furthermore, a two-level multi-view iterative optimization is designed to generate robust learning targets for refining individual views' representation learning. Theoretical analysis reveals that MVCAN works by achieving the multi-view consistency, complementarity, and noise robustness. Finally, experiments on extensive public datasets demonstrate that MVCAN outperforms state-of-the-art methods and is robust against the existence of noisy views.

Similarity-based representation learning has shown impressive capabilities in both supervised (e.g., metric learning) and unsupervised (e.g., contrastive learning) scenarios. Existing approaches effectively constrained the representation difference (i.e., the disagreement between the embeddings of two instances) to fit the corresponding (pseudo) similarity supervision. However, most of them can hardly restrict the variation of representation difference, sometimes leading to overfitting results where the clusters are disordered by drastically changed differences. In this paper, we thus propose a novel difference alignment regularization (DAR) to encourage all representation differences between inter-class instances to be as close as possible, so that the learning algorithm can produce consistent differences to distinguish data points from each other. To this end, we construct a new cross-total-variation (CTV) norm to measure the divergence among representation differences, and we convert it into an equivalent stochastic form for easy optimization. Then, we integrate the proposed regularizer into the empirical loss for difference-aligned similarity learning (DASL), shrinking the hypothesis space and alleviating overfitting. Theoretically, we prove that our regularizer tightens the error bound of the traditional similarity learning. Experiments on multi-domain data demonstrate the superiority of DASL over existing approaches in both supervised metric learning and unsupervised contrastive learning tasks.

Recent years have witnessed a burgeoning interest in federated learning (FL). However, the contexts in which clients engage in sequential learning remain under- explored. Bridging FL and continual learning (CL) gives rise to a challenging practical problem: federated continual learning (FCL). Existing research in FCL primarily focuses on mitigating the catastrophic forgetting issue of continual learning while collaborating with other clients. We argue that forgetting phenomena are not invariably detrimental. In this paper, we consider a more practical and challenging FCL setting characterized by potentially unrelated or even antagonistic data/tasks across different clients. In the FL scenario, statistical heterogeneity and data noise among clients may exhibit spurious correlations which result in biased feature learning. While existing CL strategies focus on the complete utilization of previous knowledge, we found that forgetting biased information was beneficial in our study. Therefore, we propose a new concept accurate forgetting (AF) and develop a novel generative-replay method AF-FCL that selectively utilizes previous knowledge in federated networks. We employ a probabilistic framework based on a normalizing flow model to quantify the credibility of previous knowledge. Comprehensive experiments affirm the superiority of our method over baselines.

In federated learning (FL), the multi-step update and data heterogeneity among clients often lead to a loss landscape with sharper minima, degenerating the performance of the resulted global model. Prevalent federated approaches incorporate sharpness-aware minimization (SAM) into local training to mitigate this problem. However, the local loss landscapes may not accurately reflect the flatness of global loss landscape in heterogeneous environments; as a result, minimizing local sharpness and calculating perturbations on client data might not align the efficacy of SAM in FL with centralized training. To overcome this challenge, we propose FedLESAM, a novel algorithm that locally estimates the direction of global perturbation on client side as the difference between global models received in the previous active and current rounds. Besides the improved quality, FedLESAM also speed up federated SAM-based approaches since it only performs once backpropagation in each iteration. Theoretically, we prove a slightly tighter bound than its original FedSAM by ensuring consistent perturbation. Empirically, we conduct comprehensive experiments on four federated benchmark datasets under three partition strategies to demonstrate the superior performance and efficiency of FedLESAM.

Complementary-label learning is a weakly supervised learning problem in which each training example is associated with one or multiple complementary labels indicating the classes to which it does not belong. Existing consistent approaches have relied on the uniform distribution assumption to model the generation of complementary labels, or on an ordinary-label training set to estimate the transition matrix in non-uniform cases. However, either condition may not be satisfied in real-world scenarios. In this paper, we propose a novel consistent approach that does not rely on these conditions. Inspired by the positive-unlabeled (PU) learning literature, we propose an unbiased risk estimator based on the Selected-Completely-at-Random assumption for complementary-label learning. We then introduce a risk-correction approach to address overfitting problems. Furthermore, we find that complementary-label learning can be expressed as a set of negative-unlabeled binary classification problems when using the one-versus-rest strategy. Extensive experimental results on both synthetic and real-world benchmark datasets validate the superiority of our proposed approach over state-of-the-art methods.

The key to multi-label image classification (MLC) is to improve model performance by leveraging label correlations. Unfortunately, it has been shown that overemphasizing co-occurrence relationships can cause the overfitting issue of the model, ultimately leading to performance degradation. In this paper, we provide a causal inference framework to show that the correlative features caused by the target object and its co-occurring objects can be regarded as a mediator, which has both positive and negative impacts on model predictions. On the positive side, the mediator enhances the recognition performance of the model by capturing co-occurrence relationships; on the negative side, it has the harmful causal effect that causes the model to make an incorrect prediction for the target object, even when only co-occurring objects are present in an image. To address this problem, we propose a counterfactual reasoning method to measure the total direct effect, achieved by enhancing the direct effect caused only by the target object. Due to the unknown location of the target object, we propose patching-based training and inference to accomplish this goal, which divides an image into multiple patches and identifies the pivot patch that contains the target object. Experimental results on multiple benchmark datasets with diverse configurations validate that the proposed method can achieve state-of-the-art performance.

In mobile and IoT systems, Federated Learning (FL) is increasingly important for effectively using data while maintaining user privacy. One key challenge in FL is managing statistical heterogeneity, such as non-i.i.d. data, arising from numerous clients and diverse data sources. This requires strategic cooperation, often with clients having similar characteristics. However, we are interested in a fundamental question: does achieving optimal cooperation necessarily entail cooperating with the most similar clients? Typically, significant model performance improvements are often realized not by partnering with the most similar models, but through leveraging complementary data. Our theoretical and empirical analyses suggest that optimal cooperation is achieved by enhancing complementarity in feature distribution while restricting the disparity in the correlation between features and targets. Accordingly, we introduce a novel framework, FedSaC, which balances similarity and complementarity in FL cooperation. Our framework aims to approximate an optimal cooperation network for each client by optimizing a weighted sum of model similarity and feature complementarity. The strength of FedSaC lies in its adaptability to various levels of data heterogeneity and multimodal scenarios. Our comprehensive unimodal and multimodal experiments demonstrate that FedSaC markedly surpasses other state-of-the-art FL methods.

To improve the ability of the large language model (LLMs) to tackle complex reasoning problems, chain-of-thoughts (CoT) methods were proposed to guide LLMs to reason step-by-step, enabling problem solving from simple to complex. State-of-the-art methods for generating such a chain involve interactive collaboration, where the learner generates candidate intermediate thoughts, evaluated by the LLM, guiding the generation of subsequent thoughts. However, a widespread yet understudied problem is that the evaluation from the LLM is typically noisy and unreliable, potentially misleading the generation process in selecting promising intermediate thoughts. In this paper, motivated by Vapnik's principle, we use pairwise-comparison evaluation instead of point-wise scoring to search for promising intermediate thoughts with the noisy feedback from the LLM. In each round, we randomly pair intermediate thoughts and directly prompt the LLM to select the more promising one from each pair, allowing us to identify the most promising thoughts through an iterative process. To further alleviate the noise in the comparison, we incorporate techniques from ensemble learning and dueling bandits, proposing two variants of the algorithm. Experiments on three real-world tasks demonstrate the effectiveness of our proposed algorithm and verify the rationale of the pairwise comparison mechanism.

The spurious correlation between the background features of the image and its label arises due to that the samples labeled with the same class in the training set often co-occurs with a specific background, which will cause the encoder to extract non-semantic features for classification, resulting in poor out-of-distribution generalization performance. Although many studies have been proposed to address this challenge, the semantic and spurious features are still difficult to accurately decouple from the original image and fail to achieve high performance with deep learning models. This paper proposes a novel perspective inspired by neural collapse to solve the spurious correlation problem through the alternate execution of environment partitioning and learning semantic masks. Specifically, we propose to assign an environment to each sample by learning a local model for each environment and using maximum likelihood probability. At the same time, we require that the learned semantic mask neurally collapses to the same simplex equiangular tight frame (ETF) in each environment after being applied to the original input. We conduct extensive experiments on four datasets, and the results demonstrate that our method significantly improves out-of-distribution performance.

A partial label (PL) specifies a set of candidate labels for an instance and partial-label learning (PLL) trains multi-class classifiers with PLs. Recently, many methods that incorporate techniques from other domains have shown strong potential. The expectation that stronger techniques would enhance performance has resulted in prominent PLL methods becoming not only highly complicated but also quite different from one another, making it challenging to choose the best direction for future algorithm design. While it is exciting to see higher performance, this leaves open a fundamental question: what makes a PLL method effective? We present a comprehensive empirical analysis of this question and summarize the success of PLL so far into some minimal algorithm design principles. Our findings reveal that high accuracy on benchmark-simulated datasets with PLs can misleadingly amplify the perceived effectiveness of some general techniques, which may improve representation learning but have limited impact on addressing the inherent challenges of PLs. We further identify the common behavior among successful PLL methods as a progressive transition from uniform to one-hot pseudo-labels, highlighting the critical role of mini-batch PL purification in achieving top performance. Based on our findings, we introduce a minimal working algorithm that is surprisingly simple yet effective, and propose an improved strategy to implement the design principles, suggesting a promising direction for improvements in PLL.

Partial-label learning (PLL) utilizes instances with PLs, where a PL includes several candidate labels but only one is the true label (TL). In PLL, identification-based strategy (IBS) purifies each PL on the fly to select the (most likely) TL for training; average-based strategy (ABS) treats all candidate labels equally for training and let trained models be able to predict TL. Although PLL research has focused on IBS for better performance, ABS is also worthy of study since modern IBS behaves like ABS in the beginning of training to prepare for PL purification and TL selection. In this paper, we analyze why ABS was unsatisfactory and propose how to improve it. Theoretically, we propose two problem settings of PLL and prove that average PL losses (APLLs) with bounded multi-class losses are always robust, while APLLs with unbounded losses may be non-robust, which is the first robustness analysis for PLL. Experimentally, we have two promising findings: ABS using bounded losses can match/exceed state-of-the-art performance of IBS using unbounded losses; after using robust APLLs to warm start, IBS can further improve upon itself. Our work draws attention to ABS research, which can in turn boost IBS and push forward the whole PLL.

Partial label learning (PLL) is an important problem that allows each training example to be labeled with a coarse candidate set with the ground-truth label included. However, in a more practical but challenging scenario, the annotator may miss the ground-truth and provide a wrong candidate set, which is known as the noisy PLL problem. To remedy this problem, we propose the PiCO+ framework that simultaneously disambiguates the candidate sets and mitigates label noise. Core to PiCO+, we develop a novel label disambiguation algorithm PiCO that consists of a contrastive learning module along with a novel class prototype-based disambiguation method. Theoretically, we show that these two components are mutually beneficial, and can be rigorously justified from an expectation-maximization (EM) algorithm perspective. To handle label noise, we extend PiCO to PiCO+, which further performs distance-based clean sample selection, and learns robust classifiers by a semi-supervised contrastive learning algorithm. Beyond this, we further investigate the robustness of PiCO+ in the context of out-of-distribution noise and incorporate a novel energy-based rejection method for improved robustness. Extensive experiments demonstrate that our proposed methods significantly outperform the current state-of-the-art approaches in standard and noisy PLL tasks and even achieve comparable results to fully supervised learning.

Although adversarial training (AT) is regarded as a potential defense against backdoor attacks, AT and its variants have only yielded unsatisfactory results or have even inversely strengthened backdoor attacks. The large discrepancy between expectations and reality motivates us to thoroughly evaluate the effectiveness of AT against backdoor attacks across various settings for AT and backdoor attacks. We find that the type and budget of perturbations used in AT are important, and AT with common perturbations is only effective for certain backdoor trigger patterns. Based on these empirical findings, we present some practical suggestions for backdoor defense, including relaxed adversarial perturbation and composite AT. This work not only boosts our confidence in AT's ability to defend against backdoor attacks but also provides some important insights for future research.

Under partial-label learning (PLL) where, for each training instance, only a set of ambiguous candidate labels containing the unknown true label is accessible, contrastive learning has recently boosted the performance of PLL on vision tasks, attributed to representations learned by contrasting the same/different classes of entities. Without access to true labels, positive points are predicted using pseudolabels that are inherently noisy, and negative points often require large batches or momentum encoders, resulting in unreliable similarity information and a high computational overhead. In this paper, we rethink a state-of-the-art contrastive PLL method PiCO [24], inspiring the design of a simple framework termed PaPi (Partial-label learning with a guided Prototypical classifier), which demonstrates significant scope for improvement in representation learning, thus contributing to label disambiguation. PaPi guides the optimization of a prototypical classifier by a linear classifier with which they share the same feature encoder, thus explicitly encouraging the representation to reflect visual similarity between categories. It is also technically appealing, as PaPi requires only a few components in PiCO with the opposite direction of guidance, and directly eliminates the contrastive learning module that would introduce noise and consume computational resources. We empirically demonstrate that PaPi significantly outperforms other PLL methods on various image classification tasks.

Knowledge distillation aims to learn a lightweight student network from a pre-trained teacher network. In practice, existing knowledge distillation methods are usually infeasible when the original training data is unavailable due to some privacy issues and data management considerations. Therefore, data-free knowledge distillation approaches proposed to collect training instances from the Internet. However, most of them have ignored the common distribution shift between the instances from original training data and webly collected data, affecting the reliability of the trained student network. To solve this problem, we propose a novel method dubbed "Knowledge Distillation between Different Distributions" (KD3), which consists of three components. Specifically, we first dynamically select useful training instances from the webly collected data according to the combined predictions of teacher network and student network. Subsequently, we align both the weighted features and classifier parameters of the two networks for knowledge memorization. Meanwhile, we also build a new contrastive learning block called MixDistribution to generate perturbed data with a new distribution for instance alignment, so that the student network can further learn a distribution-invariant representation. Intensive experiments on various benchmark datasets demonstrate that our proposed KD3can outperform the state-of-the-art data-free knowledge distillation approaches.

Existing knowledge distillation methods typically work by imparting the knowledge of output logits or intermediate feature maps from the teacher network to the student network, which is very successful in multi-class single-label learning. However, these methods can hardly be extended to the multi-label learning scenario, where each instance is associated with multiple semantic labels, because the prediction probabilities do not sum to one and feature maps of the whole example may ignore minor classes in such a scenario. In this paper, we propose a novel multi-label knowledge distillation method. On one hand, it exploits the informative semantic knowledge from the logits by dividing the multi-label learning problem into a set of binary classification problems; on the other hand, it enhances the distinctiveness of the learned feature representations by leveraging the structural information of label-wise embeddings. Experimental results on multiple benchmark datasets validate that the proposed method can avoid knowledge counteraction among labels, thus achieving superior performance against diverse comparing methods. Our code is available at: https://github.com/penghui-yang/L2D.

Generating unlabeled data has been recently shown to help address the few-shot hypothesis adaptation (FHA) problem, where we aim to train a classifier for the target domain with a few labeled target-domain data and a well-trained source-domain classifier (i.e., a source hypothesis), for the additional information of the highly-compatible unlabeled data. However, the generated data of the existing methods are extremely similar or even the same. The strong dependency among the generated data will lead the learning to fail. In this paper, we propose a diversity-enhancing generative network (DEG-Net) for the FHA problem, which can generate diverse unlabeled data with the help of a kernel independence measure: the Hilbert-Schmidt independence criterion (HSIC). Specifically, DEG-Net will generate data via minimizing the HSIC value (i.e., maximizing the independence) among the semantic features of the generated data. By DEG-Net, the generated unlabeled data are more diverse and more effective for addressing the FHA problem. Experimental results show that the DEG-Net outperforms existing FHA baselines and further verifies that generating diverse data plays an important role in addressing the FHA problem.

In conventional supervised classification, true labels are required for individual instances. However, it could be prohibitive to collect the true labels for individual instances, due to privacy concerns or unaffordable annotation costs. This motivates the study on classification from aggregate observations (CFAO), where the supervision is provided to groups of instances, instead of individual instances. CFAO is a generalized learning framework that contains various learning problems, such as multiple-instance learning and learning from label proportions. The goal of this paper is to present a novel universal method of CFAO, which holds an unbiased estimator of the classification risk for arbitrary losses—previous research failed to achieve this goal. Practically, our method works by weighing the importance of each instance and each label in the group, which provides purified supervision for the classifier to learn. Theoretically, our proposed method not only guarantees the risk consistency due to the unbiased risk estimator but also can be compatible with arbitrary losses. Extensive experiments on various problems of CFAO demonstrate the superiority of our proposed method.

In the presence of noisy labels, designing robust loss functions is critical for securing the generalization performance of deep neural networks. Cross Entropy (CE) loss has been shown to be not robust to noisy labels due to its unboundedness. To alleviate this issue, existing works typically design specialized robust losses with the symmetric condition, which usually lead to the underfitting issue. In this paper, our key idea is to induce a loss bound at the logit level, thus universally enhancing the noise robustness of existing losses. Specifically, we propose logit clipping (LogitClip), which clamps the norm of the logit vector to ensure that it is upper bounded by a constant. In this manner, CE loss equipped with our LogitClip method is effectively bounded, mitigating the overfitting to examples with noisy labels. Moreover, we present theoretical analyses to certify the noise-tolerant ability of LogitClip. Extensive experiments show that LogitClip not only significantly improves the noise robustness of CE loss, but also broadly enhances the generalization performance of popular robust losses.

Distribution shift (DS) may have two levels: the distribution itself changes, and the support (i.e., the set where the probability density is non-zero) also changes. When considering the support change between the training and test distributions, there can be four cases: (i) they exactly match; (ii) the training support is wider (and thus covers the test support); (iii) the test support is wider; (iv) they partially overlap. Existing methods are good at cases (i) and (ii), while cases (iii) and (iv) are more common nowadays but still under-explored. In this paper, we generalize importance weighting (IW), a golden solver for cases (i) and (ii), to a universal solver for all cases. Specifically, we first investigate why IW might fail in cases (iii) and (iv); based on the findings, we propose generalized IW (GIW) that could handle cases (iii) and (iv) and would reduce to IW in cases (i) and (ii). In GIW, the test support is split into an in-training (IT) part and an out-of-training (OOT) part, and the expected risk is decomposed into a weighted classification term over the IT part and a standard classification term over the OOT part, which guarantees the risk consistency of GIW. Then, the implementation of GIW consists of three components: (a) the split of validation data is carried out by the one-class support vector machine, (b) the first term of the empirical risk can be handled by any IW algorithm given training data and IT validation data, and (c) the second term just involves OOT validation data. Experiments demonstrate that GIW is a universal solver for DS problems, outperforming IW methods in cases (iii) and (iv).

Recently, learning with soft labels has been shown to achieve better performance than learning with hard labels in terms of model generalization, calibration, and robustness. However, collecting pointwise labeling confidence for all training examples can be challenging and time-consuming in real-world scenarios. This paper delves into a novel weakly supervised binary classification problem called confidence-difference (ConfDiff) classification. Instead of pointwise labeling confidence, we are given only unlabeled data pairs with confidence difference that specifies the difference in the probabilities of being positive. We propose a risk-consistent approach to tackle this problem and show that the estimation error bound achieves the optimal convergence rate. We also introduce a risk correction approach to mitigate overfitting problems, whose consistency and convergence rate are also proven. Extensive experiments on benchmark data sets and a real-world recommender system data set validate the effectiveness of our proposed approaches in exploiting the supervision information of the confidence difference.

Pseudo-labeling has emerged as a popular and effective approach for utilizing unlabeled data. However, in the context of semi-supervised multi-label learning (SSMLL), conventional pseudo-labeling methods encounter difficulties when dealing with instances associated with multiple labels and an unknown label count. These limitations often result in the introduction of false positive labels or the neglect of true positive ones. To overcome these challenges, this paper proposes a novel solution called Class-Aware Pseudo-Labeling (CAP) that performs pseudo-labeling in a class-aware manner. The proposed approach introduces a regularized learning framework incorporating class-aware thresholds, which effectively control the assignment of positive and negative pseudo-labels for each class. Notably, even with a small proportion of labeled examples, our observations demonstrate that the estimated class distribution serves as a reliable approximation. Motivated by this finding, we develop a class-distribution-aware thresholding strategy to ensure the alignment of pseudo-label distribution with the true distribution. The correctness of the estimated class distribution is theoretically verified, and a generalization error bound is provided for our proposed method. Extensive experiments on multiple benchmark datasets confirm the efficacy of CAP in addressing the challenges of SSMLL problems.

Recently, numerous studies have demonstrated the effectiveness of contrastive learning (CL), which learns feature representations by pulling in positive samples while pushing away negative samples. Many successes of CL lie in that there exists semantic consistency between data augmentations of the same instance. In multi-view scenarios, however, CL might cause representation degeneration when the collected multiple views inherently have inconsistent semantic information or their representations subsequently do not capture sufficient discriminative information. To address this issue, we propose a novel framework called SEM: SElf-weighted Multi-view contrastive learning with reconstruction regularization. Specifically, SEM is a general framework where we propose to first measure the discrepancy between pairwise representations and then minimize the corresponding self-weighted contrastive loss, and thus making SEM adaptively strengthen the useful pairwise views and also weaken the unreliable pairwise views. Meanwhile, we impose a self-supervised reconstruction term to regularize the hidden features of encoders, to assist CL in accessing sufficient discriminative information of data. Experiments on public multi-view datasets verified that SEM can mitigate representation degeneration in existing CL methods and help them achieve significant performance improvements. Ablation studies also demonstrated the effectiveness of SEM with different options of weighting strategies and reconstruction terms.

Out-of-distribution (OOD) detection is important for deploying reliable machine learning models on real-world applications. Recent advances in outlier exposure have shown promising results on OOD detection via fine-tuning model with informatively sampled auxiliary outliers. However, previous methods assume that the collected outliers can be sufficiently large and representative to cover the boundary between ID and OOD data, which might be impractical and challenging. In this work, we propose a novel framework, namely, Diversified Outlier Exposure (DivOE), for effective OOD detection via informative extrapolation based on the given auxiliary outliers. Specifically, DivOE introduces a new learning objective, which diversifies the auxiliary distribution by explicitly synthesizing more informative outliers for extrapolation during training. It leverages a multi-step optimization method to generate novel outliers beyond the original ones, which is compatible with many variants of outlier exposure. Extensive experiments and analyses have been conducted to characterize and demonstrate the effectiveness of the proposed DivOE. The code is publicly available at: https://github.com/tmlr-group/DivOE.

Deep reinforcement learning (DRL) provides an agent with an optimal policy so as to maximize the cumulative rewards. The policy defined in DRL mainly depends on the state, historical memory, and policy model parameters. However, we humans usually take actions according to our own intentions, such as moving fast or slow, besides the elements included in the traditional policy models. In order to make the action-choosing mechanism more similar to humans and make the agent to select actions that incorporate intentions, we propose an intention-aware policy learning method in this letter To formalize this process, we first define an intention-aware policy by incorporating the intention information into the policy model, which is learned by maximizing the cumulative rewards with the mutual information (MI) between the intention and the action. Then we derive an approximation of the MI objective that can be optimized efficiently. Finally, we demonstrate the effectiveness of the intention-aware policy in the classical MuJoCo control task and the multigoal continuous chain walking task.

Label noise is ubiquitous in many real-world scenarios which often misleads training algorithm and brings about the degraded classification performance. Therefore, many approaches have been proposed to correct the loss function given corrupted labels to combat such label noise. Among them, a trend of works achieve this goal by unbiasedly estimating the data centroid, which plays an important role in constructing an unbiased risk estimator for minimization. However, they usually handle the noisy labels in different classes all at once, so the local information inherited by each class is ignored which often leads to unsatisfactory performance. To address this defect, this paper presents a novel robust learning algorithm dubbed "Class-Wise Denoising" (CWD), which tackles the noisy labels in a class-wise way to ease the entire noise correction task. Specifically, two virtual auxiliary sets are respectively constructed by presuming that the positive and negative labels in the training set are clean, so the original false-negative labels and false-positive ones are tackled separately. As a result, an improved centroid estimator can be designed which helps to yield more accurate risk estimator. Theoretically, we prove that: 1) the variance in centroid estimation can often be reduced by our CWD when compared with existing methods with unbiased centroid estimator; and 2) the performance of CWD trained on the noisy set will converge to that of the optimal classifier trained on the clean set with a convergence rate [Formula: see text] where n is the number of the training examples. These sound theoretical properties critically enable our CWD to produce the improved classification performance under label noise, which is also demonstrated by the comparisons with ten representative state-of-the-art methods on a variety of benchmark datasets.

In label-noise learning, estimating the transition matrix is a hot topic as the matrix plays an important role in building statistically consistent classifiers. Traditionally, the transition from clean labels to noisy labels (i.e., clean-label transition matrix (CLTM)) has been widely exploited on class-dependent label-noise (wherein all samples in a clean class share the same label transition matrix). However, the CLTM cannot handle the more common instance-dependent label-noise well (wherein the clean-to-noisy label transition matrix needs to be estimated at the instance level by considering the input quality). Motivated by the fact that classifiers mostly output Bayes optimal labels for prediction, in this paper, we study to directly model the transition from Bayes optimal labels to noisy labels (i.e., Bayes-Label Transition Matrix (BLTM)) and learn a classifier to predict Bayes optimal labels. Note that given only noisy data, it is ill-posed to estimate either the CLTM or the BLTM. But favorably, Bayes optimal labels have no uncertainty compared with the clean labels, i.e., the class posteriors of Bayes optimal labels are one-hot vectors while those of clean labels are not. This enables two advantages to estimate the BLTM, i.e., (a) a set of examples with theoretically guaranteed Bayes optimal labels can be collected out of noisy data; (b) the feasible solution space is much smaller. By exploiting the advantages, this work proposes a parametrical model for estimating the instance-dependent label-noise transition matrix by employing a deep neural network, leading to better generalization and superior classification performance.

Inmultiple-instance learning(MIL), each training example is represented by a bag of instances. A training bag is either negative if it contains no positive instances or positive if it has at least one positive instance. Previous MIL methods generally assume that training bags are fully labeled. However, the exact labels of training examples may not be accessible, due to security, confidentiality, and privacy concerns. Fortunately, it could be easier for us to access the pairwise similarity between two bags (indicating whether two bags share the same label or not) and unlabeled bags, as we do not need to know the underlying label of each bag. In this paper, we provide the first attempt to investigate MIL from only similar-dissimilar-unlabeled bags. To solve this new MIL problem, we first propose a strong baseline method that trains an instance-level classifier by employing an unlabeled-unlabeled learning strategy. Then, we also propose to train a bag-level classifier based on a convex formulation and theoretically derive a generalization error bound for this method. Comprehensive experimental results show that our instance-level classifier works well, while our bag-level classifier even has better performance.

''Knowledge should not be accessible only to those who can pay" said Robert May, chair of UC's faculty Academic Senate. Similarly, machine learning should not be accessible only to those who can pay. Thus, machine learning should benefit to the whole world, especially for developing countries in Africa and Asia. When dataset sizes grow bigger, it is laborious and expensive to obtain clean supervision, especially for developing countries. As a result, the volume of noisy supervision becomes enormous, e.g., web-scale image and speech data with noisy labels. However, standard machine learning assumes that the supervised information is fully clean and intact. Therefore, noisy data harms the performance of most of the standard learning algorithms, and sometimes even makes existing algorithms broken down.

In label-noise learning, estimating the transition matrix has attracted more and more attention as the matrix plays an important role in building statistically consistent classifiers. However, it is very challenging to estimate the transition matrix T(x), where x denotes the instance, because it is unidentifiable under the instance-dependent noise (IDN). To address this problem, we have noticed that, there are psychological and physiological evidences showing that we humans are more likely to annotate instances of similar appearances to the same classes, and thus poor-quality or ambiguous instances of similar appearances are easier to be mislabeled to the correlated or same noisy classes. Therefore, we propose assumption on the geometry of T(x) that “the closer two instances are, the more similar their corresponding transition matrices should be”. More specifically, we formulate above assumption into the manifold embedding, to effectively reduce the degree of freedom of T(x) and make it stably estimable in practice. The proposed manifold-regularized technique works by directly reducing the estimation error without hurting the approximation error about the estimation problem of T(x). Experimental evaluations on four synthetic and two real-world datasets demonstrate that our method is superior to state-of-the-art approaches for label-noise learning under the challenging IDN.

The AutoAttack (AA) has been the most reliable method to evaluate adversarial robustness when considerable computational resources are available. However, the high computational cost (e.g., 100 times more than that of the project gradient descent attack) makes AA infeasible for practitioners with limited computational resources, and also hinders applications of AA in the adversarial training (AT). In this paper, we propose a novel method, minimum-margin (MM) attack, to fast and reliably evaluate adversarial robustness. Compared with AA, our method achieves comparable performance but only costs 3% of the computational time in extensive experiments. The reliability of our method lies in that we evaluate the quality of adversarial examples using the margin between two targets that can precisely identify the most adversarial example. The computational efficiency of our method lies in an effective Sequential TArget Ranking Selection (STARS) method, ensuring that the cost of the MM attack is independent of the number of classes. The MM attack opens a new way for evaluating adversarial robustness and provides a feasible and reliable way to generate high-quality adversarial examples in AT.

Label smoothing (LS) is an arising learning paradigm that uses the positively weighted average of both the hard training labels and uniformly distributed soft labels. It was shown that LS serves as a regularizer for training data with hard labels and therefore improves the generalization of the model. Later it was reported LS even helps with improving robustness when learning with noisy labels. However, we observed that the advantage of LS vanishes when we operate in a high label noise regime. Intuitively speaking, this is due to the increased entropy of P(noisy label|X) when the noise rate is high, in which case, further applying LS tends to “over-smooth” the estimated posterior. We proceeded to discover that several learning-with-noisy-labels solutions in the literature instead relate more closely to negative/not label smoothing (NLS), which acts counter to LS and defines as using a negative weight to combine the hard and soft labels! We provide understandings for the properties of LS and NLS when learning with noisy labels. Among other established properties, we theoretically show NLS is considered more beneficial when the label noise rates are high. We provide extensive experimental results on multiple benchmarks to support our findings too. Code is publicly available at https://github.com/UCSC-REAL/negative-label-smoothing.

In label-noise learning, estimating the transition matrix is a hot topic as the matrix plays an important role in building statistically consistent classifiers. Traditionally, the transition from clean labels to noisy labels (i.e., clean-label transition matrix (CLTM)) has been widely exploited to learn a clean label classifier by employing the noisy data. Motivated by that classifiers mostly output Bayes optimal labels for prediction, in this paper, we study to directly model the transition from Bayes optimal labels to noisy labels (i.e., Bayes-label transition matrix (BLTM)) and learn a classifier to predict Bayes optimal labels. Note that given only noisy data, it is ill-posed to estimate either the CLTM or the BLTM. But favorably, Bayes optimal labels have less uncertainty compared with the clean labels, i.e., the class posteriors of Bayes optimal labels are one-hot vectors while those of clean labels are not. This enables two advantages to estimate the BLTM, i.e., (a) a set of examples with theoretically guaranteed Bayes optimal labels can be collected out of noisy data; (b) the feasible solution space is much smaller. By exploiting the advantages, we estimate the BLTM parametrically by employing a deep neural network, leading to better generalization and superior classification performance.

Adversarial training (AT) with imperfect supervision is significant but receives limited attention. To push AT towards more practical scenarios, we explore a brand new yet challenging setting, i.e., AT with complementary labels (CLs), which specify a class that a data sample does not belong to. However, the direct combination of AT with existing methods for CLs results in consistent failure, but not on a simple baseline of two-stage training. In this paper, we further explore the phenomenon and identify the underlying challenges of AT with CLs as intractable adversarial optimization and low-quality adversarial examples. To address the above problems, we propose a new learning strategy using gradually informative attacks, which consists of two critical components: 1) Warm-up Attack (Warm-up) gently raises the adversarial perturbation budgets to ease the adversarial optimization with CLs; 2) Pseudo-Label Attack (PLA) incorporates the progressively informative model predictions into a corrected complementary loss. Extensive experiments are conducted to demonstrate the effectiveness of our method on a range of benchmarked datasets. The code is publicly available at: https://github.com/RoyalSkye/ATCL.

Contrastive learning (CL) pretrains feature embeddings to scatter instances in the feature space so that the training data can be well discriminated. Most existing CL techniques usually encourage learning such feature embeddings in the highdimensional space to maximize the instance discrimination. However, this practice may lead to undesired results where the scattering instances are sparsely distributed in the high-dimensional feature space, making it difficult to capture the underlying similarity between pairwise instances. To this end, we propose a novel framework called contrastive learning with low-dimensional reconstruction (CLLR), which adopts a regularized projection layer to reduce the dimensionality of the feature embedding. In CLLR, we build the sparse / low-rank regularizer to adaptively reconstruct a low-dimensional projection space while preserving the basic objective for instance discrimination, and thus successfully learning contrastive embeddings that alleviate the above issue. Theoretically, we prove a tighter error bound for CLLR; empirically, the superiority of CLLR is demonstrated across multiple domains. Both theoretical and experimental results emphasize the significance of learning low-dimensional contrastive embeddings.

\emph{Classification with rejection} (CwR) refrains from making a prediction to avoid critical misclassification when encountering test samples that are difficult to classify. Though previous methods for CwR have been provided with theoretical guarantees, they are only compatible with certain loss functions, making them not flexible enough when the loss needs to be changed with the dataset in practice. In this paper, we derive a novel formulation for CwR that can be equipped with arbitrary loss functions while maintaining the theoretical guarantees. First, we show that $K$-class CwR is equivalent to a $(K\!+\!1)$-class classification problem on the original data distribution with an augmented class, and propose an empirical risk minimization formulation to solve this problem with an estimation error bound. Then, we find necessary and sufficient conditions for the learning \emph{consistency} of the surrogates constructed on our proposed formulation equipped with any classification-calibrated multi-class losses, where consistency means the surrogate risk minimization implies the target risk minimization for CwR. Finally, experiments on benchmark datasets validate the effectiveness of our proposed method.

In rank aggregation (RA), a collection of preferences from different users are summarized into a total order under the assumption of homogeneity of users. Model misspecification in RA arises since the homogeneity assumption fails to be satisfied in the complex real-world situation. Existing robust RAs usually resort to an augmentation of the ranking model to account for additional noises, where the collected preferences can be treated as a noisy perturbation of idealized preferences. Since the majority of robust RAs rely on certain perturbation assumptions, they cannot generalize well to agnostic noise-corrupted preferences in the real world. In this paper, we propose CoarsenRank, which possesses robustness against model misspecification. Specifically, the properties of our CoarsenRank are summarized as follows: (1) CoarsenRank is designed for mild model misspecification, which assumes there exist the ideal preferences (consistent with model assumption) that locate in a neighborhood of the actual preferences. (2) CoarsenRank then performs regular RAs over a neighborhood of the preferences instead of the original data set directly. Therefore, CoarsenRank enjoys robustness against model misspecification within a neighborhood. (3) The neighborhood of the data set is defined via their empirical data distributions. Further, we put an exponential prior on the unknown size of the neighborhood and derive a much-simplified posterior formula for CoarsenRank under particular divergence measures. (4) CoarsenRank is further instantiated to Coarsened Thurstone, Coarsened Bradly-Terry, and Coarsened Plackett-Luce with three popular probability ranking models. Meanwhile, tractable optimization strategies are introduced with regards to each instantiation respectively. In the end, we apply CoarsenRank on four real-world data sets. Experiments show that CoarsenRank is fast and robust, achieving consistent improvements over baseline methods.

SU classification employs similar (S) data pairs (two examples belong to the same class) and unlabeled (U) data points to build a classifier, which can serve as an alternative to the standard supervised trained classifiers requiring data points with class labels. SU classification is advantageous because in the era of big data, more attention has been paid to data privacy. Datasets with specific class labels are often difficult to obtain in real-world classification applications regarding privacy-sensitive matters, such as politics and religion, which can be a bottleneck in supervised classification. Fortunately, similarity labels do not reveal the explicit information and inherently protect the privacy, e.g., collecting answers to “With whom do you share the same opinion on issue $\mathcal{I}$?" instead of “What is your opinion on issue $\mathcal{I}$?". Nevertheless, SU classification still has an obvious limitation: respondents might answer these questions in a manner that is viewed favorably by others instead of answering truthfully. Therefore, there exist some dissimilar data pairs labeled as similar, which significantly degenerates the performance of SU classification. In this paper, we study how to learn from noisy similar (nS) data pairs and unlabeled (U) data, which is called nSU classification. Specifically, we carefully model the similarity noise and estimate the noise rate by using the mixture proportion estimation technique. Then, a clean classifier can be learned by minimizing a denoised and unbiased classification risk estimator, which only involves the noisy data. Moreover, we further derive a theoretical generalization error bound for the proposed method. Experimental results demonstrate the effectiveness of the proposed algorithm on several benchmark datasets.

The drastic increase of data quantity often brings the severe decrease of data quality, such as incorrect label annotations. It poses a great challenge for robustly training Deep Neural Networks (DNNs). Existing learning methods with label noise either employ ad-hoc heuristics or restrict to specific noise assumptions. However, more general situations, such as instance-dependent label noise, have not been fully explored, as scarce studies focus on their label corruption process. By categorizing instances into confusing and unconfusing instances, this paper proposes a simple yet universal probabilistic model, which explicitly relates noisy labels to their instances. The resultant model can be realized by DNNs, where the training procedure is accomplished by employing a novel alternating optimization algorithm. Experiments on datasets with both synthetic and real-world label noise verify the proposed method yields significant improvements on robustness over state-of-the-art counterparts.

We consider the situation in which a user has collected a small set of documents on a cohesive topic, and they want to retrieve additional documents on this topic from a large collection.Information Retrieval (IR) solutions treat the document set as a query, and look for similar documents in the collection.We propose to extend the IR approach by treating the problem as an instance of positive-unlabeled (PU) learning-i.e., learning binary classifiers from only positive (the query documents) and unlabeled (the results of the IR engine) data.Utilizing PU learning for text with big neural networks is a largely unexplored field.We discuss various challenges in applying PU learning to the setting, showing that the standard implementations of state-of-the-art PU solutions fail.We propose solutions for each of the challenges and empirically validate them with ablation tests.We demonstrate the effectiveness of the new method using a series of experiments of retrieving PubMed abstracts adhering to fine-grained topics, showing improvements over the common IR solution and other baselines.

\emph{Contrastive learning} (CL) pretrains models in a pairwise manner, where given a data point, other data points are all regarded as dissimilar, including some that are \emph{semantically} similar. The issue has been addressed by properly weighting similar and dissimilar pairs as in \emph{positive-unlabeled learning}, so that the objective of CL is \emph{unbiased} and CL is \emph{consistent}. However, in this paper, we argue that this great solution is still not enough: its weighted objective \emph{hides} the issue where the semantically similar pairs are still pushed away; as CL is pretraining, this phenomenon is not our desideratum and might affect downstream tasks. To this end, we propose \emph{large-margin contrastive learning} (LMCL) with \emph{distance polarization regularizer}, motivated by the distribution characteristic of pairwise distances in \emph{metric learning}. In LMCL, we can distinguish between \emph{intra-cluster} and \emph{inter-cluster} pairs, and then only push away inter-cluster pairs, which \emph{solves} the above issue explicitly. Theoretically, we prove a tighter error bound for LMCL; empirically, the superiority of LMCL is demonstrated across multiple domains, \emph{i.e.}, image classification, sentence representation, and reinforcement learning.

In learning with noisy labels, for every instance, its label can randomly walk to other classes following a transition distribution which is named a noise model. Well-studied noise models are all instance-independent, namely, the transition depends only on the original label but not the instance itself, and thus they are less practical in the wild. Fortunately, methods based on instance-dependent noise have been studied, but most of them have to rely on strong assumptions on the noise models. To alleviate this issue, we introduce confidence-scored instance-dependent noise (CSIDN), where each instance-label pair is equipped with a confidence score. We find that with the help of confidence scores, the transition distribution of each instance can be approximately estimated. Similarly to the powerful forward correction for instance-independent noise, we propose a novel instance-level forward correction for CSIDN. We demonstrate the utility and effectiveness of our method through multiple experiments on datasets with synthetic label noise and real-world unknown noise.

Weakly supervised learning has drawn considerable attention recently to reduce the expensive time and labor consumption of labeling massive data. In this paper, we investigate a novel weakly supervised learning problem of learning from similarity-confidence (Sconf) data, where only unlabeled data pairs equipped with confidence that illustrates their degree of similarity (two examples are similar if they belong to the same class) are needed for training a discriminative binary classifier. We propose an unbiased estimator of the classification risk that can be calculated from only Sconf data and show that the estimation error bound achieves the optimal convergence rate. To alleviate potential overfitting when flexible models are used, we further employ a risk correction scheme on the proposed risk estimator. Experimental results demonstrate the effectiveness of the proposed methods.

In adversarial training (AT), the main focus has been the objective and optimizer while the model has been less studied, so that the models being used are still those classic ones in standard training (ST). Classic network architectures (NAs) are generally worse than searched NA in ST, which should be the same in AT. In this paper, we argue that NA and AT cannot be handled independently, since given a dataset, the optimal NA in ST would be no longer optimal in AT. That being said, AT is time-consuming itself; if we directly search NAs in AT over large search spaces, the computation will be practically infeasible. Thus, we propose diverse-structured network (DS-Net), to significantly reduce the size of the search space: instead of low-level operations, we only consider predefined atomic blocks, where an atomic block is a time-tested building block like the residual block. There are only a few atomic blocks and thus we can weight all atomic blocks rather than find the best one in a searched block of DS-Net, which is an essential tradeoff between exploring diverse structures and exploiting the best structures. Empirical results demonstrate the advantages of DS-Net, i.e., weighting the atomic blocks.

To alleviate the data requirement for training effective binary classifiers in binary classification, many weakly supervised learning settings have been proposed. Among them, some consider using pairwise but not pointwise labels, when pointwise labels are not accessible due to privacy, confidentiality, or security reasons. However, as a pairwise label denotes whether or not two data points share a pointwise label, it cannot be easily collected if either point is equally likely to be positive or negative. Thus, in this paper, we propose a novel setting called pairwise comparison (Pcomp) classification, where we have only pairs of unlabeled data that we know one is more likely to be positive than the other. Firstly, we give a Pcomp data generation process, derive an unbiased risk estimator (URE) with theoretical guarantee, and further improve URE using correction functions. Secondly, we link Pcomp classification to noisy-label learning to develop a progressive URE and improve it by imposing consistency regularization. Finally, we demonstrate by experiments the effectiveness of our methods, which suggests Pcomp is a valuable and practically useful type of pairwise supervision besides the pairwise label.

The maximum mean discrepancy (MMD) test could in principle detect any distributional discrepancy between two datasets. However, it has been shown that the MMD test is unaware of adversarial attacks–the MMD test failed to detect the discrepancy between natural data and adversarial data. Given this phenomenon, we raise a question: are natural and adversarial data really from different distributions? The answer is affirmative–the previous use of the MMD test on the purpose missed three key factors, and accordingly, we propose three components. Firstly, the Gaussian kernel has limited representation power, and we replace it with an effective deep kernel. Secondly, the test power of the MMD test was neglected, and we maximize it following asymptotic statistics. Finally, adversarial data may be non-independent, and we overcome this issue with the help of wild bootstrap. By taking care of the three factors, we verify that the MMD test is aware of adversarial attacks, which lights up a novel road for adversarial data detection based on two-sample tests.

In label-noise learning, the transition matrix plays a key role in building statistically consistent classifiers. Existing consistent estimators for the transition matrix have been developed by exploiting anchor points. However, the anchor-point assumption is not always satisfied in real scenarios. In this paper, we propose an end-to-end framework for solving label-noise learning without anchor points, in which we simultaneously optimize two objectives: the cross entropy loss between the noisy label and the predicted probability by the neural network, and the volume of the simplex formed by the columns of the transition matrix. Our proposed framework can identify the transition matrix if the clean class-posterior probabilities are sufficiently scattered. This is by far the mildest assumption under which the transition matrix is provably identifiable and the learned classifier is statistically consistent. Experimental results on benchmark datasets demonstrate the effectiveness and robustness of the proposed method.

To cope with high annotation costs, training a classifier only from weakly supervised data has attracted a great deal of attention these days. Among various approaches, strengthening supervision from completely unsupervised classification is a promising direction, which typically employs class priors as the only supervision and trains a binary classifier from unlabeled (U) datasets. While existing risk-consistent methods are theoretically grounded with high flexibility, they can learn only from two U sets. In this paper, we propose a new approach for binary classification from $m$ U-sets for $m\ge2$. Our key idea is to consider an auxiliary classification task called surrogate set classification (SSC), which is aimed at predicting from which U set each observed sample is drawn. SSC can be solved by a standard (multi-class) classification method, and we use the SSC solution to obtain the final binary classifier through a certain linear-fractional transformation. We built our method in a flexible and efficient end-to-end deep learning framework and prove it to be classifier-consistent. Through experiments, we demonstrate the superiority of our proposed method over state-of-the-art methods.

Learning with noisy labels has attracted a lot of attention in recent years, where the mainstream approaches are in \emph{pointwise} manners. Meanwhile, \emph{pairwise} manners have shown great potential in supervised metric learning and unsupervised contrastive learning. Thus, a natural question is raised: does learning in a pairwise manner \emph{mitigate} label noise? To give an affirmative answer, in this paper, we propose a framework called \emph{Class2Simi}: it transforms data points with noisy \emph{class labels} to data pairs with noisy \emph{similarity labels}, where a similarity label denotes whether a pair shares the class label or not. Through this transformation, the \emph{reduction of the noise rate} is theoretically guaranteed, and hence it is in principle easier to handle noisy similarity labels. Amazingly, DNNs that predict the \emph{clean} class labels can be trained from noisy data pairs if they are first pretrained from noisy data points. Class2Simi is \emph{computationally efficient} because not only this transformation is on-the-fly in mini-batches, but also it just changes loss computation on top of model prediction into a pairwise manner. Its effectiveness is verified by extensive experiments.

We investigate the adversarial robustness of CNNs from the perspective of channel-wise activations. By comparing normally trained and adversarially trained models, we observe that adversarial training (AT) robustifies CNNs by aligning the channel-wise activations of adversarial data with those of their natural counterparts. However, the channels that are \textit{negatively-relevant} (NR) to predictions are still over-activated when processing adversarial data. Besides, we also observe that AT does not result in similar robustness for all classes. For the robust classes, channels with larger activation magnitudes are usually more \textit{positively-relevant} (PR) to predictions, but this alignment does not hold for the non-robust classes. Given these observations, we hypothesize that suppressing NR channels and aligning PR ones with their relevances further enhances the robustness of CNNs under AT. To examine this hypothesis, we introduce a novel mechanism, \textit{i.e.}, \underline{C}hannel-wise \underline{I}mportance-based \underline{F}eature \underline{S}election (CIFS). The CIFS manipulates channels’ activations of certain layers by generating non-negative multipliers to these channels based on their relevances to predictions. Extensive experiments on benchmark datasets including CIFAR10 and SVHN clearly verify the hypothesis and CIFS’s effectiveness of robustifying CNNs.

Many weakly supervised classification methods employ a noise transition matrix to capture the class-conditional label corruption. To estimate the transition matrix from noisy data, existing methods often need to estimate the noisy class-posterior, which could be unreliable due to the overconfidence of neural networks. In this work, we propose a theoretically grounded method that can estimate the noise transition matrix and learn a classifier simultaneously, without relying on the error-prone noisy class-posterior estimation. Concretely, inspired by the characteristics of the stochastic label corruption process, we propose total variation regularization, which encourages the predicted probabilities to be more distinguishable from each other. Under mild assumptions, the proposed method yields a consistent estimator of the transition matrix. We show the effectiveness of the proposed method through experiments on benchmark and real-world datasets.

Multiple-instance learning (MIL) is an important weakly supervised binary classification problem, where training instances are arranged in bags, and each bag is assigned a positive or negative label. Most of the previous studies for MIL assume that training bags are fully labeled. However, in some real-world scenarios, it could be difficult to collect fully labeled bags, due to the expensive time and labor consumption of the labeling task. Fortunately, it could be much easier for us to collect similar and dissimilar bags (indicating whether two bags share the same label or not), because we do not need to figure out the underlying label of each bag in this case. Therefore, in this paper, we for the first time investigate MIL from only similar and dissimilar bags. To solve this new MIL problem, we propose a convex formulation to train a bag-level classifier based on empirical risk minimization and theoretically derive a generalization error bound. In addition, we also propose a strong baseline for this new MIL problem, which aims to train an instance-level classifier by minimizing the instance-level empirical risk. Extensive experimental results clearly demonstrate that our proposed baseline works well, while our proposed convex formulation is even better.

The memorization effect of deep neural network (DNN) plays a pivotal role in many state-of-the-art label-noise learning methods. To exploit this property, the early stopping trick, which stops the optimization at the early stage of training, is usually adopted. Current methods generally decide the early stopping point by considering a DNN as a whole. However, a DNN can be considered as a composition of a series of layers, and we find that the latter layers in a DNN are much more sensitive to label noise, while their former counterparts are quite robust. Therefore, selecting a stopping point for the whole network may make different DNN layers antagonistically affect each other, thus degrading the final performance. In this paper, we propose to separate a DNN into different parts and progressively train them to address this problem. Instead of the early stopping which trains a whole DNN all at once, we initially train former DNN layers by optimizing the DNN with a relatively large number of epochs. During training, we progressively train the latter DNN layers by using a smaller number of epochs with the preceding layers fixed to counteract the impact of noisy labels. We term the proposed method as progressive early stopping (PES). Despite its simplicity, compared with the traditional early stopping, PES can help to obtain more promising and stable results. Furthermore, by combining PES with existing approaches on noisy label training, we achieve state-of-the-art performance on image classification benchmarks. The code is made public at https://github.com/tmllab/PES.

Reweighting adversarial data during training has been recently shown to improve adversarial robustness, where data closer to the current decision boundaries are regarded as more critical and given larger weights. However, existing methods measuring the closeness are not very reliable: they are discrete and can take only a few values, and they are path-dependent, i.e., they may change given the same start and end points with different attack paths. In this paper, we propose three types of probabilistic margin (PM), which are continuous and path-independent, for measuring the aforementioned closeness and reweighing adversarial data. Specifically, a PM is defined as the difference between two estimated class-posterior probabilities, e.g., such a probability of the true label minus the probability of the most confusing label given some natural data. Though different PMs capture different geometric properties, all three PMs share a negative correlation with the vulnerability of data: data with larger/smaller PMs are safer/riskier and should have smaller/larger weights. Experiments demonstrated that PMs are reliable and PM-based reweighting methods outperformed state-of-the-art counterparts.

Label noise generally degenerates the performance of deep learning algorithms because deep neural networks easily overfit label errors. Let $X$ and $Y$ denote the instance and clean label, respectively. When $Y$ is a cause of $X$, according to which many datasets have been constructed, e.g., \textit{SVHN} and \textit{CIFAR}, the distributions of $P(X)$ and $P(Y|X)$ are generally entangled. This means that the unsupervised instances are helpful to learn the classifier and thus reduce the side effect of label noise. However, it remains elusive on how to exploit the causal information to handle the label-noise problem. We propose to model and make use of the causal process in order to correct the label-noise effect.Empirically, the proposed method outperforms all state-of-the-art methods on both synthetic and real-world label-noise datasets.

Recent advances in weakly supervised classification allow us to train a classifier from only positive and unlabeled (PU) data. However, existing PU classification methods typically require an accurate estimate of the class-prior probability, a critical bottleneck particularly for high-dimensional data. This problem has been commonly addressed by applying principal component analysis in advance, but such unsupervised dimension reduction can collapse the underlying class structure. In this letter, we propose a novel representation learning method from PU data based on the information-maximization principle. Our method does not require class-prior estimation and thus can be used as a preprocessing method for PU classification. Through experiments, we demonstrate that our method, combined with deep neural networks, highly improves the accuracy of PU class-prior estimation, leading to state-of-the-art PU classification performance.

Summarizing large-scale directed graphs into small-scale representations is a useful but less-studied problem setting. Conventional clustering approaches, based on Min-Cut-style criteria, compress both the vertices and edges of the graph into the communities, which lead to a loss of directed edge information. On the other hand, compressing the vertices while preserving the directed-edge information provides a way to learn the small-scale representation of a directed graph. The reconstruction error, which measures the edge information preserved by the summarized graph, can be used to learn such representation. Compared to the original graphs, the summarized graphs are easier to analyze and are capable of extracting group-level features, useful for efficient interventions of population behavior. In this letter, we present a model, based on minimizing reconstruction error with nonnegative constraints, which relates to a Max-Cut criterion that simultaneously identifies the compressed nodes and the directed compressed relations between these nodes. A multiplicative update algorithm with column-wise normalization is proposed. We further provide theoretical results on the identifiability of the model and the convergence of the proposed algorithms. Experiments are conducted to demonstrate the accuracy and robustness of the proposed method.

Exact recovery of tensor decomposition (TD) methods is a desirable property in both unsupervised learning and scientific data analysis. The numerical defects of TD methods, however, limit their practical applications on real-world data. As an alternative, convex tensor decomposition (CTD) was proposed to alleviate these problems, but its exact-recovery property is not properly addressed so far. To this end, we focus on latent convex tensor decomposition (LCTD), a practically widely-used CTD model, and rigorously prove a sufficient condition for its exact-recovery property. Furthermore, we show that such property can be also achieved by a more general model than LCTD. In the new model, we generalize the classic tensor (un-)folding into reshuffling operation, a more flexible mapping to relocate the entries of the matrix into a tensor. Armed with the reshuffling operations and exact-recovery property, we explore a totally novel application for (generalized) LCTD, i.e., image steganography. Experimental results on synthetic data validate our theory, and results on image steganography show that our method outperforms the state-of-the-art methods.

The recently proposed unlabeled-unlabeled (UU) classification method allows us to train a binary classifier only from two unlabeled datasets with different class priors. Since this method is based on the empirical risk minimization, it works as if it is a supervised classification method, compatible with any model and optimizer. However, this method sometimes suffers from severe overfitting, which we would like to prevent in this paper. Our empirical finding in applying the original UU method is that overfitting often co-occurs with the empirical risk going negative, which is not legitimate. Therefore, we propose to wrap the terms that cause a negative empirical risk by certain correction functions. Then, we prove the consistency of the corrected risk estimator and derive an estimation error bound for the corrected risk minimizer. Experiments show that our proposal can successfully mitigate overfitting of the UU method and significantly improve the classification accuracy.

Graph embedding has shown its effectiveness to represent graph information and capture deep relationships in graph data. Most recent graph embedding methods focus on attributed graphs, since they preserve both structure and content information in the network. However, corruption can exist in the graph structure as well as the node content of the graph, and both can lead to inferior embedding results. Unfortunately, few existing graph embedding algorithms have considered the corruption problem, and to the best of our knowledge, none has studied structural corruption in attributed graphs, including missing and redundant edges. This field is difficult for previous methods, mainly due to two challenges: (1) the existence of various corruption causes has made it difficult to recognize corruptions in graphs, and (2) the complexity of graph-structured data has increased the difficulty of handling corruption therein for graph embedding methods. These facts lead us here to propose a novel autoencoder-based graph embedding approach, which is robust against structural corruption. Our idea comes from the recent discovery of memorization effects in deep learning. Namely, deep neural networks prefer to fit clean data first, before they over-fit corrupted data. Specifically, we train two autoencoders simultaneously and let them learn the reliability of the edges in the graph from each other. The two autoencoders would evaluate the edges according to their reconstructed structure and manipulate this by devaluing those distrusted edges to update the structure information. The updated structure would be used further in the next iteration as the ground-truth of its peer-network. Experiments on different versions of real-world graphs show state-of-the-art results and demonstrate the robustness of our model against structural corruption.

Given data with noisy labels, over-parameterized deep networks can gradually memorize the data, and fit everything in the end. Although equipped with corrections for noisy labels, many learning methods in this area still suffer overfitting due to undesired memorization. In this paper, to relieve this issue, we propose stochastic integrated gradient underweighted ascent (SIGUA): in a mini-batch, we adopt gradient descent on good data as usual, and learning-rate-reduced gradient ascent on bad data; the proposal is a versatile approach where data goodness or badness is w.r.t. desired or undesired memorization given a base learning method. Technically, SIGUA pulls optimization back for generalization when their goals conflict with each other; philosophically, SIGUA shows forgetting undesired memorization can reinforce desired memorization. Experiments demonstrate that SIGUA successfully robustifies two typical base learning methods, so that their performance is often significantly improved.

In weakly supervised learning, unbiased risk estimator(URE) is a powerful tool for training classifiers when training and test data are drawn from different distributions. Nevertheless, UREs lead to overfitting in many problem settings when the models are complex like deep networks. In this paper, we investigate reasons for such overfitting by studying a weakly supervised problem called learning with complementary labels. We argue the quality of gradient estimation matters more in risk minimization. Theoretically, we show that a URE gives an unbiased gradient estimator(UGE). Practically, however, UGEs may suffer from huge variance, which causes empirical gradients to be usually far away from true gradients during minimization. To this end, we propose a novel surrogate complementary loss(SCL) framework that trades zero bias with reduced variance and makes empirical gradients more aligned with true gradients in the direction. Thanks to this characteristic, SCL successfully mitigates the overfitting issue and improves URE-based methods.

A complementary label (CL) simply indicates an incorrect class of an example, but learning with CLs results in multi-class classifiers that can predict the correct class. Unfortunately, the problem setting only allows a single CL for each example, which notably limits its potential since our labelers may easily identify multiple CLs (MCLs) to one example. In this paper, we propose a novel problem setting to allow MCLs for each example and two ways for learning with MCLs. In the first way, we design two wrappers that decompose MCLs into many single CLs, so that we could use any method for learning with CLs. However, the supervision information that MCLs hold is conceptually diluted after decomposition. Thus, in the second way, we derive an unbiased risk estimator; minimizing it processes each set of MCLs as a whole and possesses an estimation error bound. We further improve the second way into minimizing properly chosen upper bounds. Experiments show that the former way works well for learning with MCLs but the latter is even better.

Overparameterized deep networks have the capacity to memorize training data with zero \emph{training error}. Even after memorization, the \emph{training loss} continues to approach zero, making the model overconfident and the test performance degraded. Since existing regularizers do not directly aim to avoid zero training loss, it is hard to tune their hyperparameters in order to maintain a fixed/preset level of training loss. We propose a direct solution called \emph{flooding} that intentionally prevents further reduction of the training loss when it reaches a reasonably small value, which we call the \emph{flood level}. Our approach makes the loss float around the flood level by doing mini-batched gradient descent as usual but gradient ascent if the training loss is below the flood level. This can be implemented with one line of code and is compatible with any stochastic optimizer and other regularizers. With flooding, the model will continue to “random walk” with the same non-zero training loss, and we expect it to drift into an area with a flat loss landscape that leads to better generalization. We experimentally show that flooding improves performance and, as a byproduct, induces a double descent curve of the test loss.

Partial-label learning (PLL) is a typical weakly supervised learning problem, where each training instance is equipped with a set of candidate labels among which only one is the true label. Most existing methods elaborately designed learning objectives as constrained optimizations that must be solved in specific manners, making their computational complexity a bottleneck for scaling up to big data. The goal of this paper is to propose a novel framework of PLL with flexibility on the model and optimization algorithm. More specifically, we propose a novel estimator of the classification risk, theoretically analyze the classifier-consistency, and establish an estimation error bound. Then we propose a progressive identification algorithm for approximately minimizing the proposed risk estimator, where the update of the model and identification of true labels are conducted in a seamless manner. The resulting algorithm is model-independent and loss-independent, and compatible with stochastic optimization. Thorough experiments demonstrate it sets the new state of the art.

Sample selection approaches are popular in robust learning from noisy labels. However, how to properly control the selection process so that deep networks can benefit from the memorization effect is a hard problem. In this paper, motivated by the success of automated machine learning (AutoML), we model this issue as a function approximation problem. Specifically, we design a domain-specific search space based on general patterns of the memorization effect and propose a novel Newton algorithm to solve the bi-level optimization problem efficiently. We further provide a theoretical analysis of the algorithm, which ensures a good approximation to critical points. Experiments are performed on both benchmark and real-world data sets. Results demonstrate that the proposed method is much better than the state-of-the-art noisy-label-learning approaches, and also much more efficient than existing AutoML algorithms.

Adversarial training based on the minimax formulation is necessary for obtaining adversarial robustness of trained models. However, it is conservative or even pessimistic so that it sometimes hurts the natural generalization. In this paper, we raise a fundamental question{—}do we have to trade off natural generalization for adversarial robustness? We argue that adversarial training is to employ confident adversarial data for updating the current model. We propose a novel formulation of friendly adversarial training (FAT): rather than employing most adversarial data maximizing the loss, we search for least adversarial data (i.e., friendly adversarial data) minimizing the loss, among the adversarial data that are confidently misclassified. Our novel formulation is easy to implement by just stopping the most adversarial data searching algorithms such as PGD (projected gradient descent) early, which we call early-stopped PGD. Theoretically, FAT is justified by an upper bound of the adversarial risk. Empirically, early-stopped PGD allows us to answer the earlier question negatively{—}adversarial robustness can indeed be achieved without compromising the natural generalization.

Under distribution shift (DS) where the training data distribution differs from the test one, a powerful technique is importance weighting (IW) which handles DS in two separate steps: weight estimation (WE) estimates the test-over-training density ratio and weighted classification (WC) trains the classifier from weighted training data. However, IW cannot work well on complex data, since WE is incompatible with deep learning. In this paper, we rethink IW and theoretically show it suffers from a circular dependency: we need not only WE for WC, but also WC for WE where a trained deep classifier is used as the feature extractor (FE). To cut off the dependency, we try to pretrain FE from unweighted training data, which leads to biased FE. To overcome the bias, we propose an end-to-end solution dynamic IW that iterates between WE and WC and combines them in a seamless manner, and hence our WE can also enjoy deep networks and stochastic optimizers indirectly. Experiments with two representative types of DS on three popular datasets show that our dynamic IW compares favorably with state-of-the-art methods.

Partial-label learning (PLL) is a multi-class classification problem, where each training example is associated with a set of candidate labels. Even though many practical PLL methods have been proposed in the last two decades, there lacks a theoretical understanding of the consistency of those methods - none of the PLL methods hitherto possesses a generation process of candidate label sets, and then it is still unclear why such a method works on a specific dataset and when it may fail given a different dataset. In this paper, we propose the first generation model of candidate label sets, and develop two PLL methods that are guaranteed to be provably consistent, i.e., one is risk-consistent and the other is classifier-consistent. Our methods are advantageous, since they are compatible with any deep network or stochastic optimizer. Furthermore, thanks to the generation model, we would be able to answer the two questions above by testing if the generation model matches given candidate label sets. Experiments on benchmark and real-world datasets validate the effectiveness of the proposed generation model and two PLL methods.

Learning with the \textit{instance-dependent} label noise is challenging, because it is hard to model such real-world noise. Note that there are psychological and physiological evidences showing that we humans perceive instances by decomposing them into parts. Annotators are therefore more likely to annotate instances based on the parts rather than the whole instances, where a wrong mapping from parts to classes may cause the instance-dependent label noise. Motivated by this human cognition, in this paper, we approximate the instance-dependent label noise by exploiting \textit{part-dependent} label noise. Specifically, since instances can be approximately reconstructed by a combination of parts, we approximate the instance-dependent \textit{transition matrix} for an instance by a combination of the transition matrices for the parts of the instance. The transition matrices for parts can be learned by exploiting anchor points (i.e., data points that belong to a specific class almost surely). Empirical evaluations on synthetic and real-world datasets demonstrate our method is superior to the state-of-the-art approaches for learning from the instance-dependent label noise.

The transition matrix, denoting the transition relationship from clean labels to noisy labels, is essential to build statistically consistent classifiers in label-noise learning. Existing methods for estimating the transition matrix rely heavily on estimating the noisy class posterior. However, the estimation error for noisy class posterior could be large because of the randomness of label noise. The estimation error would lead the transition matrix to be poorly estimated. Therefore in this paper, we aim to solve this problem by exploiting the divide-and-conquer paradigm. Specifically, we introduce an intermediate class to avoid directly estimating the noisy class posterior. By this intermediate class, the original transition matrix can then be factorized into the product of two easy-to-estimated transition matrices. We term the proposed method as the dual $T$-estimator. Both theoretical analyses and empirical results illustrate the effectiveness of the dual $T$-estimator for estimating transition matrices, leading to better classification performances.

In binary classification, there are situations where negative (N) data are too diverse to be fully labeled and we often resort to positive-unlabeled (PU) learning in these scenarios. However, collecting a non-representative N set that contains only a small portion of all possible N data can often be much easier in practice. This paper studies a novel classification framework which incorporates such biased N (bN) data in PU learning. We provide a method based on empirical risk minimization to address this PUbN classification problem. Our approach can be regarded as a novel example-weighting algorithm, with the weight of each example computed through a preliminary step that draws inspiration from PU learning. We also derive an estimation error bound for the proposed method. Experimental results demonstrate the effectiveness of our algorithm in not only PUbN learning scenarios but also ordinary PU learning scenarios on several benchmark datasets.

In contrast to the standard classification paradigm where the true class is given to each training pattern, complementary-label learning only uses training patterns each equipped with a complementary label, which only specifies one of the classes that the pattern does not belong to. The goal of this paper is to derive a novel framework of complementary-label learning with an unbiased estimator of the classification risk, for arbitrary losses and models—all existing methods have failed to achieve this goal. Not only is this beneficial for the learning stage, it also makes model/hyper-parameter selection (through cross-validation) possible without the need of any ordinarily labeled validation data, while using any linear/non-linear models or convex/non-convex loss functions. We further improve the risk estimator by a non-negative correction and gradient ascent trick, and demonstrate its superiority through experiments.

Learning with noisy labels is one of the hottest problems in weakly-supervised learning. Based on memorization effects of deep neural networks, training on small-loss instances becomes very promising for handling noisy labels. This fosters the state-of-the-art approach "Co-teaching" that cross-trains two deep neural networks using the small-loss trick. However, with the increase of epochs, two networks converge to a consensus and Co-teaching reduces to the self-training MentorNet. To tackle this issue, we propose a robust learning paradigm called Co-teaching+, which bridges the "Update by Disagreement” strategy with the original Co-teaching. First, two networks feed forward and predict all data, but keep prediction disagreement data only. Then, among such disagreement data, each network selects its small-loss data, but back propagates the small-loss data from its peer network and updates its own parameters. Empirical results on benchmark datasets demonstrate that Co-teaching+ is much superior to many state-of-the-art methods in the robustness of trained models.

In label-noise learning, the \textit{noise transition matrix}, denoting the probabilities that clean labels flip into noisy labels, plays a central role in building \textit{statistically consistent classifiers}. Existing theories have shown that the transition matrix can be learned by exploiting \textit{anchor points} (i.e., data points that belong to a specific class almost surely). However, when there are no anchor points, the transition matrix will be poorly learned, and those previously consistent classifiers will significantly degenerate. In this paper, without employing anchor points, we propose a \textit{transition-revision} ($T$-Revision) method to effectively learn transition matrices, leading to better classifiers. Specifically, to learn a transition matrix, we first initialize it by exploiting data points that are similar to anchor points, having high \textit{noisy class posterior probabilities}. Then, we modify the initialized matrix by adding a \textit{slack variable}, which can be learned and validated together with the classifier by using noisy data. Empirical results on benchmark-simulated and real-world label-noise datasets demonstrate that without using exact anchor points, the proposed method is superior to state-of-the-art label-noise learning methods.

Uncoupled regression is the problem to learn a model from unlabeled data and the set of target values while the correspondence between them is unknown. Such a situation arises in predicting anonymized targets that involve sensitive information, e.g., one's annual income. Since existing methods for uncoupled regression often require strong assumptions on the true target function, and thus, their range of applications is limited, we introduce a novel framework that does not require such assumptions in this paper. Our key idea is to utilize \emph{pairwise comparison data, which consists of pairs of unlabeled data that we know which one has a larger target value. Such pairwise comparison data is easy to collect, as typically discussed in the learning-to-rank scenario, and does not break the anonymity of data. We propose two practical methods for uncoupled regression from pairwise comparison data and show that the learned regression model converges to the optimal model with the optimal parametric convergence rate when the target variable distributes uniformly. Moreover, we empirically show that for linear models the proposed methods are comparable to ordinary supervised regression with labeled data.

Supervised learning needs a huge amount of labeled data, which can be a big bottleneck under the situation where there is a privacy concern or labeling cost is high. To overcome this problem, we propose a new weakly-supervised learning setting where only similar (S) data pairs (two examples belong to the same class) and unlabeled (U) data points are needed instead of fully labeled data, which is called SU classification. We show that an unbiased estimator of the classification risk can be obtained only from SU data, and the estimation error of its empirical risk minimizer achieves the optimal parametric convergence rate. Finally, we demonstrate the effectiveness of the proposed method through experiments.

Distributionally Robust Supervised Learning (DRSL) is necessary for building reliable machine learning systems. When machine learning is deployed in the real world, its performance can be significantly degraded because test data may follow a different distribution from training data. DRSL with f-divergences explicitly considers the worst-case distribution shift by minimizing the adversarially reweighted training loss. In this paper, we analyze this DRSL, focusing on the classification scenario. Since the DRSL is explicitly formulated for a distribution shift scenario, we naturally expect it to give a robust classifier that can aggressively handle shifted distributions. However, surprisingly, we prove that the DRSL just ends up giving a classifier that exactly fits the given training distribution, which is too pessimistic. This pessimism comes from two sources: the particular losses used in classification and the fact that the variety of distributions to which the DRSL tries to be robust is too wide. Motivated by our analysis, we propose simple DRSL that overcomes this pessimism and empirically demonstrate its effectiveness.

Feature missing is a serious problem in many applications, which may lead to low quality of training data and further significantly degrade the learning performance. While feature acquisition usually involves special devices or complex process, it is expensive to acquire all feature values for the whole dataset. On the other hand, features may be correlated with each other, and some values may be recovered from the others. It is thus important to decide which features are most informative for recovering the other features as well as improving the learning performance. In this paper, we try to train an effective classification model with least acquisition cost by jointly performing active feature querying and supervised matrix completion. When completing the feature matrix, a novel objective function is proposed to simultaneously minimize the reconstruction error on observed entries and the supervised loss on training data. When querying the feature value, the most uncertain entry is actively selected based on the variance of previous iterations. In addition, a bi-objective optimization method is presented for cost-aware active selection when features bear different acquisition costs. The effectiveness of the proposed approach is well validated by both theoretical analysis and experimental study.

It is important to learn various types of classifiers given training data with noisy labels. Noisy labels, in the most popular noise model hitherto, are corrupted from ground-truth labels by an unknown noise transition matrix. Thus, by estimating this matrix, classifiers can escape from overfitting those noisy labels. However, such estimation is practically difficult, due to either the indirect nature of two-step approaches, or not big enough data to afford end-to-end approaches. In this paper, we propose a human-assisted approach called ''Masking'' that conveys human cognition of invalid class transitions and naturally speculates the structure of the noise transition matrix. To this end, we derive a structure-aware probabilistic model incorporating a structure prior, and solve the challenges from structure extraction and structure alignment. Thanks to Masking, we only estimate unmasked noise transition probabilities and the burden of estimation is tremendously reduced. We conduct extensive experiments on CIFAR-10 and CIFAR-100 with three noise structures as well as the industrial-level Clothing1M with agnostic noise structure, and the results show that Masking can improve the robustness of classifiers significantly.

Deep learning with noisy labels is practically challenging, as the capacity of deep models is so high that they can totally memorize these noisy labels sooner or later during training. Nonetheless, recent studies on the memorization effects of deep neural networks show that they would first memorize training data of clean labels and then those of noisy labels. Therefore in this paper, we propose a new deep learning paradigm called ''Co-teaching'' for combating with noisy labels. Namely, we train two deep neural networks simultaneously, and let them teach each other given every mini-batch: firstly, each network feeds forward all data and selects some data of possibly clean labels; secondly, two networks communicate with each other what data in this mini-batch should be used for training; finally, each network back propagates the data selected by its peer network and updates itself. Empirical results on noisy versions of MNIST, CIFAR-10 and CIFAR-100 demonstrate that Co-teaching is much superior to the state-of-the-art methods in the robustness of trained deep models.

Can we learn a binary classifier from only positive data, without any negative data or unlabeled data? We show that if one can equip positive data with confidence (positive-confidence), one can successfully learn a binary classifier, which we name positive-confidence (Pconf) classification. Our work is related to one-class classification which is aimed at "describing" the positive class by clustering-related methods, but one-class classification does not have the ability to tune hyper-parameters and their aim is not on "discriminating" positive and negative classes. For the Pconf classification problem, we provide a simple empirical risk minimization framework that is model-independent and optimization-independent. We theoretically establish the consistency and an estimation error bound, and demonstrate the usefulness of the proposed method for training deep neural networks through experiments.

Sufficient dimension reduction (SDR) is aimed at obtaining the low-rank projection matrix in the input space such that information about output data is maximally preserved. Among various approaches to SDR, a promising method is based on the eigendecomposition of the outer product of the gradient of the conditional density of output given input. In this letter, we propose a novel estimator of the gradient of the logarithmic conditional density that directly fits a linear-in-parameter model to the true gradient under the squared loss. Thanks to this simple least-squares formulation, its solution can be computed efficiently in a closed form. Then we develop a new SDR method based on the proposed gradient estimator. We theoretically prove that the proposed gradient estimator, as well as the SDR solution obtained from it, achieves the optimal parametric convergence rate. Finally, we experimentally demonstrate that our SDR method compares favorably with existing approaches in both accuracy and computational efficiency on a variety of artificial and benchmark data sets.

\emphNon-Gaussian component analysis (NGCA) is an unsupervised linear dimension reduction method that extracts low-dimensional non-Gaussian “signals” from high-dimensional data contaminated with Gaussian noise. NGCA can be regarded as a generalization of \emphprojection pursuit (PP) and \emphindependent component analysis (ICA) to multi-dimensional and dependent non-Gaussian components. Indeed, seminal approaches to NGCA are based on PP and ICA. Recently, a novel NGCA approach called \emphleast-squares NGCA (LSNGCA) has been developed, which gives a solution analytically through least-squares estimation of \emphlog-density gradients and eigendecomposition. However, since \emphpre-whitening of data is involved in LSNGCA, it performs unreliably when the data covariance matrix is ill-conditioned, which is often the case in high-dimensional data analysis. In this paper, we propose a \emphwhitening-free variant of LSNGCA and experimentally demonstrate its superiority.

Most of the semi-supervised classification methods developed so far use unlabeled data for regularization purposes under particular distributional assumptions such as the cluster assumption. In contrast, recently developed methods of classification from positive and unlabeled data (PU classification) use unlabeled data for risk evaluation, i.e., label information is directly extracted from unlabeled data. In this paper, we extend PU classification to also incorporate negative data and propose a novel semi-supervised learning approach. We establish generalization error bounds for our novel methods and show that the bounds decrease with respect to the number of unlabeled data without the distributional assumptions that are required in existing semi-supervised learning methods. Through experiments, we demonstrate the usefulness of the proposed methods.

Collecting labeled data is costly and thus a critical bottleneck in real-world classification tasks. To mitigate this problem, we propose a novel setting, namely learning from complementary labels for multi-class classification. A complementary label specifies a class that a pattern does not belong to. Collecting complementary labels would be less laborious than collecting ordinary labels, since users do not have to carefully choose the correct class from a long list of candidate classes. However, complementary labels are less informative than ordinary labels and thus a suitable approach is needed to better learn from them. In this paper, we show that an unbiased estimator to the classification risk can be obtained only from complementarily labeled data, if a loss function satisfies a particular symmetric condition. We derive estimation error bounds for the proposed method and prove that the optimal parametric convergence rate is achieved. We further show that learning from complementary labels can be easily combined with learning from ordinary labels (i.e., ordinary supervised learning), providing a highly practical implementation of the proposed method. Finally, we experimentally demonstrate the usefulness of the proposed methods.

From only positive (P) and unlabeled (U) data, a binary classifier could be trained with PU learning, in which the state of the art is unbiased PU learning. However, if its model is very flexible, empirical risks on training data will go negative, and we will suffer from serious overfitting. In this paper, we propose a non-negative risk estimator for PU learning: when getting minimized, it is more robust against overfitting, and thus we are able to use very flexible models (such as deep neural networks) given limited P data. Moreover, we analyze the bias, consistency, and mean-squared-error reduction of the proposed risk estimator, and bound the estimation error of the resulting empirical risk minimizer. Experiments demonstrate that our risk estimator fixes the overfitting problem of its unbiased counterparts.

Non-Gaussian component analysis (NGCA) is aimed at identifying a linear subspace such that the projected data follows a non-Gaussian distribution. In this paper, we propose a novel NGCA algorithm based on log-density gradient estimation. Unlike existing methods, the proposed NGCA algorithm identifies the linear subspace by using the eigenvalue decomposition without any iterative procedures, and thus is computationally reasonable. Furthermore, through theoretical analysis, we prove that the identified subspace converges to the true subspace at the optimal parametric rate. Finally, the practical performance of the proposed algorithm is demonstrated on both artificial and benchmark datasets.

In PU learning, a binary classifier is trained from positive (P) and unlabeled (U) data without negative (N) data. Although N data is missing, it sometimes outperforms PN learning (i.e., ordinary supervised learning). Hitherto, neither theoretical nor experimental analysis has been given to explain this phenomenon. In this paper, we theoretically compare PU (and NU) learning against PN learning based on the upper bounds on estimation errors. We find simple conditions when PU and NU learning are likely to outperform PN learning, and we prove that, in terms of the upper bounds, either PU or NU learning (depending on the class-prior probability and the sizes of P and N data) given infinite U data will improve on PN learning. Our theoretical findings well agree with the experimental results on artificial and benchmark data even when the experimental setup does not match the theoretical assumptions exactly.

Estimating the derivatives of probability density functions is an essential step in statistical data analysis. A naive approach to estimate the derivatives is to first perform density estimation and then compute its derivatives. However, this approach can be unreliable because a good density estimator does not necessarily mean a good density derivative estimator. To cope with this problem, in this letter, we propose a novel method that directly estimates density derivatives without going through density estimation. The proposed method provides computationally efficient estimation for the derivatives of any order on multidimensional data with a hyperparameter tuning method and achieves the optimal parametric convergence rate. We further discuss an extension of the proposed method by applying regularized multitask learning and a general framework for density derivative estimation based on Bregman divergences. Applications of the proposed method to nonparametric Kullback-Leibler divergence approximation and bandwidth matrix selection in kernel density estimation are also explored.

We discuss binary classification from only from positive and unlabeled data (PU classification), which is conceivable in various real-world machine learning problems. Since unlabeled data consists of both positive and negative data, simply separating positive and unlabeled data yields a biased solution. Recently, it was shown that the bias can be canceled by using a particular non-convex loss such as the ramp loss. However, classifier training with a non-convex loss is not straightforward in practice. In this paper, we discuss a convex formulation for PU classification that can still cancel the bias. The key idea is to use different loss functions for positive and unlabeled samples. However, in this setup, the hinge loss is not permissible. As an alternative, we propose the double hinge loss. Theoretically, we prove that the estimators converge to the optimal solutions at the optimal parametric rate. Experimentally, we demonstrate that PU classification with the double hinge loss performs as accurate as the non-convex method, with a much lower computational cost.

Given a hypothesis space, the large volume principle by Vladimir Vapnik prioritizes equivalence classes according to their volume in the hypothesis space. The volume approximation has hitherto been successfully applied to binary learning problems. In this paper, we propose a novel generalization to multiple classes, allowing applications of the large volume principle on more learning problems such as multi-class, multi-label and serendipitous learning in a transductive manner. Although the resultant learning method involves a non-convex optimization problem, the globally optimal solution is almost surely unique and can be obtained using O(n^3) time. Novel theoretical analyses are presented for the proposed method, and experimental results show it compares favorably with the one-vs-rest extension.

We propose a general information-theoretic approach to semi-supervised metric learning called SERAPH (SEmi-supervised metRic leArning Paradigm with Hypersparsity) that does not rely on the manifold assumption. Given the probability parameterized by a Mahalanobis distance, we maximize its entropy on labeled data and minimize its entropy on unlabeled data following entropy regularization. For metric learning, entropy regularization improves manifold regularization by considering the dissimilarity information of unlabeled data in the unsupervised part, and hence it allows the supervised and unsupervised parts to be integrated in a natural and meaningful way. Moreover, we regularize SERAPH by trace-norm regularization to encourage low-dimensional projections associated with the distance metric. The nonconvex optimization problem of SERAPH could be solved efficiently and stably by either a gradient projection algorithm or an EM-like iterative algorithm whose M-step is convex. Experiments demonstrate that SERAPH compares favorably with many well-known metric learning methods, and the learned Mahalanobis distance possesses high discriminability even under noisy environments.

Information-maximization clustering learns a probabilistic classifier in an unsupervised manner so that mutual information between feature vectors and cluster assignments is maximized. A notable advantage of this approach is that it involves only continuous optimization of model parameters, which is substantially simpler than discrete optimization of cluster assignments. However, existing methods still involve nonconvex optimization problems, and therefore finding a good local optimal solution is not straightforward in practice. In this letter, we propose an alternative information-maximization clustering method based on a squared-loss variant of mutual information. This novel approach gives a clustering solution analytically in a computationally efficient way via kernel eigenvalue decomposition. Furthermore, we provide a practical model selection procedure that allows us to objectively optimize tuning parameters included in the kernel function. Through experiments, we demonstrate the usefulness of the proposed approach.

We propose squared-loss mutual information regularization (SMIR) for multi-class probabilistic classification, following the information maximization principle. SMIR is convex under mild conditions and thus improves the nonconvexity of mutual information regularization. It offers all of the following four abilities to semi-supervised algorithms: Analytical solution, out-of-sample/multi-class classification, and probabilistic output. Furthermore, novel generalization error bounds are derived. Experiments show SMIR compares favorably with state-of-the-art methods.

Policy gradient is a useful model-free reinforcement learning approach, but it tends to suffer from instability of gradient estimates. In this paper, we analyze and improve the stability of policy gradient methods. We first prove that the variance of gradient estimates in the PGPE(policy gradients with parameter-based exploration) method is smaller than that of the classical REINFORCE method under a mild assumption. We then derive the optimal baseline for PGPE, which contributes to further reducing the variance. We also theoretically show that PGPE with the optimal baseline is more preferable than REINFORCE with the optimal baseline in terms of the variance of gradient estimates. Finally, we demonstrate the usefulness of the improved PGPE method through experiments.

Most well-known discriminative clustering models, such as spectral clustering (SC) and maximum margin clustering (MMC), are non-Bayesian. Moreover, they merely considered to embed domain-dependent prior knowledge into data-specific kernels, while other forms of prior knowledge were seldom considered in these models. In this paper, we propose a Bayesian maximum margin clustering model (BMMC) based on the low-density separation assumption, which unifies the merits of both Bayesian and discriminative approaches. In addition to stating prior distribution on functions explicitly as traditional Gaussian processes, special prior knowledge can be embedded into BMMC implicitly via the Universum set easily. Furthermore, it is much easier to solve a BMMC than an MMC since the integer variables in the optimization are eliminated. Experimental results show that the BMMC achieves comparable or even better performance than state-of-the-art clustering methods and solving BMMC is more efficiently.