Accurately annotating large scale dataset is notoriously expensive both in time and in money. Although acquiring low-quality-annotated dataset can be much cheaper, it often badly damages the performance of trained models when using such dataset without particular treatment. Various methods have been proposed for learning with noisy labels. However, most methods only handle limited kinds of noise patterns, require auxiliary information or steps (e.g. , knowing or estimating the noise transition matrix), or lack theoretical justification. In this paper, we propose a novel information-theoretic loss function, $\mathcal{L}_{DMI}$, for training deep neural networks robust to label noise. The core of $\mathcal{L}_{DMI}$ is a generalized version of mutual information, termed Determinant based Mutual Information (DMI), which is not only information-monotone but also relatively invariant. \emph{To the best of our knowledge, $\mathcal{L}_{DMI}$ is the first loss function that is provably robust to instance-independent label noise, regardless of noise pattern, and it can be applied to any existing classification neural networks straightforwardly without any auxiliary information}. In addition to theoretical justification, we also empirically show that using $\mathcal{L}_{DMI}$ outperforms all other counterparts in the classification task on both image dataset and natural language dataset include Fashion-MNIST, CIFAR-10, Dogs vs. Cats, MR with a variety of synthesized noise patterns and noise amounts, as well as a real-world dataset Clothing1M. Codes are available at https://github.com/Newbeeer/L_DMI .
Information-Theoretic Decomposition for Multimodal Interaction Learning
Gaoling School of Artificial Intelligence, Renmin University of China, Beijing（中国人民大学人工智能学院，北京）；Beijing Key Laboratory of Research on Large Models（北京大模型研究关键实验室）；Engineering Research Center of Next-Generation Intelligent Search（下一代智能搜索与推荐工程研究中心）；Beihang University, Beijing（北航，北京）；Gaotu Techedu Inc.（高图科技有限公司）
Idiap Research Institute（Idiap研究所）；École Polytechnique Fédéral de Lausanne（洛桑联邦理工学院）
Collecting labeled data to train deep neural networks is costly and even impractical for many tasks. Thus, research effort has been focused in automatically curated datasets or unsupervised and weakly supervised learning. The common problem in these directions is learning with unreliable label information. In this paper, we address the tolerance of deep embedding learning losses against label noise, i.e. when the observed labels are different from the true labels. Specifically, we provide the sufficient conditions to achieve theoretical guarantees for the 2 common loss functions: marginal loss…
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；Inception Institute of Artificial Intelligence（起源人工智能研究院）
Theoretical analysis of the divide-and-conquer based distributed learning with least square loss in the reproducing kernel Hilbert space (RKHS) have recently been explored within the framework of learning theory. However, the studies on learning theory for general loss functions and hypothesis spaces remain limited. To fill the gap, we study the risk performance of distributed empirical risk minimization (ERM) for general loss functions and hypothesis spaces. The main contributions are two-fold. First, we derive two tight risk bounds under certain basic assumptions on the hypothesis space, as …
In this paper, we present and illustrate some new tools for rigorously analyzing training data selection methods. These tools focus on the information theoretic losses that occur when sampling data. We use this framework to prove that two methods, Facility Location Selection and Transductive Experimental Design, reduce these losses. These are meant to act as generalizable theoretical examples of applying the field of Information Theoretic Deep Learning Theory to the fields of data selection and active learning. Both analyses yield insight into their respective methods and increase their interp…
We propose a theoretical framework for an adaptive learning rate policy for the Mean Absolute Error loss function and Quantile loss function and evaluate its effectiveness for regression tasks. The framework is based on the theory of Lipschitz continuity, specifically utilizing the relationship between learning rate and Lipschitz constant of the loss function. Based on experimentation, we have found that the adaptive learning rate policy enables up to 20x faster convergence compared to a constant learning rate policy.
National Key Laboratory for Novel Software Technology（计算机软件新技术国家重点实验室）；Nanjing University（南京大学）
Deep neural networks need large amounts of labeled data to achieve good performance. In real-world applications, labels are usually collected from non-experts such as crowdsourcing to save cost and thus are noisy. In the past few years, deep learning methods for dealing with noisy labels have been developed, many of which are based on the small-loss criterion. However, there are few theoretical analyses to explain why these methods could learn well from noisy labels. In this paper, we theoretically explain why the widely-used small-loss criterion works. Based on the explanation, we reformalize…
Shenzhen Origin AI Technology Co. Ltd（深圳元始智能科技有限公司）；Southern University of Science and Technology（南方科技大学）；University of California, Berkeley（加利福尼亚大学伯克利分校）
Generalized zero shot learning (GZSL) is still a technical challenge of deep learning as it has to recognize both source and target classes without data from target classes. To preserve the semantic relation between source and target classes when only trained with data from source classes, we address the quantification of the knowledge transfer and semantic relation from an information-theoretic viewpoint. To this end, we follow the prototypical model and format the variables of concern as a probability vector. Leveraging on the proposed probability vector representation, the information measu…
Graduate School at Shenzhen, Tsinghua University（清华大学深圳研究生院）
In this paper, we propose a new discriminative model named \emph{nonextensive information theoretical machine (NITM)} based on nonextensive generalization of Shannon information theory. In NITM, weight parameters are treated as random variables. Tsallis divergence is used to regularize the distribution of weight parameters and maximum unnormalized Tsallis entropy distribution is used to evaluate fitting effect. On the one hand, it is showed that some well-known margin-based loss functions such as $\ell_{0/1}$ loss, hinge loss, squared hinge loss and exponential loss can be unified by unnormali…
This paper proposes a deep representation learning using an information-theoretic loss with an aim to increase the inter-class distances as well as within-class similarity in the embedded space. Tasks such as anomaly and out-of-distribution detection, in which test samples comes from classes unseen in training, are problematic for deep neural networks. For such tasks, it is not sufficient to merely discriminate between known classes. Our intuition is to represent the known classes in compact and separated embedded regions in order to decrease the possibility of known and unseen classes overlap…
New York University（纽约大学）；Shanghai Jiao Tong University（上海交通大学）；Fundan University（复旦大学）
This paper studies distributed estimation and support recovery for high-dimensional linear regression model with heavy-tailed noise. To deal with heavy-tailed noise whose variance can be infinite, we adopt the quantile regression loss function instead of the commonly used squared loss. However, the non-smooth quantile loss poses new challenges to high-dimensional distributed estimation in both computation and theoretical development. To address the challenge, we transform the response variable and establish a new connection between quantile regression and ordinary linear regression. Then, we p…
University of Twente（特文特大学）；Peking University（北京大学）；Samsung Research China-Beijing（三星中国研究院-北京）；Pazhou Lab（琶洲实验室）
As an important branch of weakly supervised learning, partial label learning deals with data where each instance is assigned with a set of candidate labels, whereas only one of them is true. Despite many methodology studies on learning from partial labels, there still lacks theoretical understandings of their risk consistent properties under relatively weak assumptions, especially on the link between theoretical results and the empirical choice of parameters. In this paper, we propose a family of loss functions named \textit{Leveraged Weighted} (LW) loss, which for the first time introduces th…
We introduce a new surrogate loss function called orbit loss in the structured prediction framework, which has good theoretical and practical advantages. While the orbit loss is not convex, it has a simple analytical gradient and a simple perceptron-like learning rule. We analyze the new loss theoretically and state a PAC-Bayesian generalization bound. We also prove that the new loss is consistent in the strong sense; namely, the risk achieved by the set of the trained parameters approaches the infimum risk achievable by any linear decoder over the given features. Methods that are aimed at ris…
The overall predictive uncertainty of a trained predictor can be decomposed into separate contributions due to epistemic and aleatoric uncertainty. Under a Bayesian formulation, assuming a well-specified model, the two contributions can be exactly expressed (for the log-loss) or bounded (for more general losses) in terms of information-theoretic quantities (Xu and Raginsky, 2020). This paper addresses the study of epistemic uncertainty within an information-theoretic framework in the broader setting of Bayesian meta-learning. A general hierarchical Bayesian model is assumed in which hyperparam…
Nara Institute of Science and Technology (NAIST)（奈良先端科学技术大学院大学）；Hokkaido University（北海道大学）
Negative sampling (NS) loss plays an important role in learning knowledge graph embedding (KGE) to handle a huge number of entities. However, the performance of KGE degrades without hyperparameters such as the margin term and number of negative samples in NS loss being appropriately selected. Currently, empirical hyperparameter tuning addresses this problem at the cost of computational time. To solve this problem, we theoretically analyzed NS loss to assist hyperparameter tuning and understand the better use of the NS loss in KGE learning. Our theoretical analysis showed that scoring methods w…
