论文检索

Theory · Domain Adaptation and Transfer Learning

Yanfang Ling, Jiyong Li, Lingbo Li, Shangsong Liang

Recent methods are proposed to improve performance of domain adaptation by inferring domain index under an adversarial variational bayesian framework, where domain index is unavailable. However, existing methods typically assume that the global domain indices are sampled from a vanilla gaussian prior, overlooking the inherent structures among different domains.To address this challenge, we propose a Bayesian Domain Adaptation with Gaussian Mixture Domain-Indexing(GMDI) algorithm. GMDI employs a Gaussian Mixture Model for domain indices, with the number of component distributions in the "domain-themes'' space adaptively determined by a Chinese Restaurant Process. By dynamically adjusting the mixtures at the domain indices level, GMDI significantly improves domain adaptation performance. Our theoretical analysis demonstrates that GMDI achieves a more stringent evidence lower bound, closer to the log-likelihood. For classification, GMDI outperforms all approaches, and surpasses the state-of-the-art method, VDI, by up to 3.4%, reaching 99.3%. For regression, GMDI reduces MSE by up to 21% (from 3.160 to 2.493), achieving the lowest errors among all methods.

Trustworthy Machine Learning

Zhe Tao, Aditya V Thakur

Ensuring that a DNN satisfies a desired property is critical when deploying DNNs in safety-critical applications. There are efficient methods that can verify whether a DNN satisfies a property, as seen in the annual DNN verification competition (VNN-COMP). However, the problem of provably editing a DNN to satisfy a property remains challenging. We present PREPARED, the first efficient technique for provable editing of DNNs. Given a DNN $\mathcal{N}$ with parameters $\theta$, input polytope $P$, and output polytope $Q$, PREPARED finds new parameters $\theta'$ such that $\forall \mathrm{x} \in P . \mathcal{N}(\mathrm{x}; \theta') \in Q$ while minimizing the changes $\lVert{\theta' - \theta}\rVert$. Given a DNN and a property it violates from the VNN-COMP benchmarks, PREPARED is able to provably edit the DNN to satisfy this property within 45 seconds. PREPARED is efficient because it relaxes the NP-hard provable editing problem to solving a linear program. The key contribution is the novel notion of Parametric Linear Relaxation, which enables PREPARED to construct tight output bounds of the DNN that are parameterized by the new parameters $\theta'$. We demonstrate that PREPARED is more efficient and effective compared to prior DNN editing approaches i) using the VNN-COMP benchmarks, ii) by editing CIFAR10 and TinyImageNet image-recognition DNNs, and BERT sentiment-classification DNNs for local robustness, and iii) by training a DNN to model a geodynamics process and satisfy physics constraints.

Theory · Domain Adaptation and Transfer Learning

Reza Ghane, Danil Akhtiamov, Babak Hassibi

We study the problem of transfer learning and fine-tuning in linear models for both regression and binary classification. In particular, we consider the use of stochastic gradient descent (SGD) on a linear model initialized with pretrained weights and using a small training data set from the target distribution. In the asymptotic regime of large models, we provide an exact and rigorous analysis and relate the generalization errors (in regression) and classification errors (in binary classification) for the pretrained and fine-tuned models. In particular, we give conditions under which the fine-tuned model outperforms the pretrained one. An important aspect of our work is that all the results are "universal", in the sense that they depend only on the first and second order statistics of the target distribution. They thus extend well beyond the standard Gaussian assumptions commonly made in the literature. Furthermore, our universality results extend beyond standard SGD training to the test error of a classification task trained using ridge regression.

Deep Learning · Everything Else

Haoyu Wang, Zhuo Huang, Zhiwei Lin, Tongliang Liu

Machine learning craves high-quality data which is a major bottleneck during realistic deployment, as it takes abundant resources and massive human labor to collect and label data. Unfortunately, label noise where image data mismatches with incorrect label exists ubiquitously in all kinds of datasets, significantly degrading the learning performance of deep networks. Learning with Label Noise (LNL) has been a common strategy for mitigating the influence of noisy labels. However, existing LNL methods either require pertaining using the memorization effect to separate clean data from noisy ones or rely on dataset assumptions that cannot extend to various scenarios. Thanks to the development of Multimodal Large Language Models (MLLMs) which possess massive knowledge and hold In-Context Learning (ICL) ability, this paper proposes NoiseGPT to effectively leverage MLLMs as a knowledge expert for conducting label noise detection and rectification. Specifically, we observe a \textit{probability curvature} effect of MLLMs where clean and noisy examples reside on curvatures with different smoothness, further enabling the detection of label noise. By designing a token-wise Mix-of-Feature (MoF) technique to produce the curvature, we propose an In-Context Discrepancy (ICD) measure to determine the authenticity of an image-label pair. Subsequently, we repeat such a process to find the best matching pairs to complete our label rectification. Through extensive experiments, we carefully demonstrate the effectiveness of NoiseGPT on detecting and cleansing dataset noise, especially on ILSVRC12, the AUROC of NoiseGPT reached over 0.92. And by integrating with existing methods, the classification performance can be significantly improved on noisy datasets, typically by 22.8\% on 80\% symmetric CIFAR-10 with M-correction. Source code: \url{https://github.com/drunkerWang/NoiseGPT}

Learning Theory

Hyunsuk Kim, Liam Hodgkinson, Ryan Theisen, Michael Mahoney

As performance gains through scaling data and/or model size experience diminishing returns, it is becoming increasingly popular to turn to ensembling, where the predictions of multiple models are combined to improve accuracy. In this paper, we provide a detailed analysis of how the disagreement and the polarization (a notion we introduce and define in this paper) among classifiers relate to the performance gain achieved by aggregating individual classifiers, for majority vote strategies in classification tasks.We address these questions in the following ways. (1) An upper bound for polarization is derived, and we propose what we call a neural polarization law: most interpolating neural network models are 4/3-polarized. Our empirical results not only support this conjecture but also show that polarization is nearly constant for a dataset, regardless of hyperparameters or architectures of classifiers. (2) The error rate of the majority vote classifier is considered under restricted entropy conditions, and we present a tight upper bound that indicates that the disagreement is linearly correlated with the error rate, and that the slope is linear in the polarization.(3) We prove results for the asymptotic behavior of the disagreement in terms of the number of classifiers, which we show can help in predicting the performance for a larger number of classifiers from that of a smaller number. Our theoretical findings are supported by empirical results on several image classification tasks with various types of neural networks.

Wei Wu, Kecheng Zheng, Shuailei Ma, Fan Lu, Yuxin Guo, Yifei Zhang, Wei Chen, Qingpei Guo, Yujun Shen, Zheng-Jun Zha

In this work, we empirically confirm that the key reason causing such an issue is that the training images are usually paired with short captions, leaving certain tokens easily overshadowed by salient tokens. Towards this problem, our initial attempt is to relabel the data with long captions, however, directly learning with which may lead to performance degradation in understanding short text (e.g., in the image classification task). Then, after incorporating corner tokens to aggregate diverse textual information, we manage to help the model catch up to its original level of short text understanding yet greatly enhance its capability of long text understanding. We further look into whether the model can continuously benefit from longer captions and notice a clear trade-off between the performance and the efficiency. Finally, we validate the effectiveness of our approach using a self-constructed large-scale dataset, which consists of 100M long caption oriented text-image pairs. Our method achieves superior performance in long-text-image retrieval tasks. The project page is available at https://wuw2019.github.io/lot-lip.

Xiaochuan Gong, Jie Hao, Mingrui Liu

This paper investigates a class of stochastic bilevel optimization problems where the upper-level function is nonconvex with potentially unbounded smoothness and the lower-level problem is strongly convex. These problems have significant applications in sequential data learning, such as text classification using recurrent neural networks. The unbounded smoothness is characterized by the smoothness constant of the upper-level function scaling linearly with the gradient norm, lacking a uniform upper bound. Existing state-of-the-art algorithms require $\widetilde{O}(\epsilon^{-4})$ oracle calls of stochastic gradient or Hessian/Jacobian-vector product to find an $\epsilon$-stationary point. However, it remains unclear if we can further improve the convergence rate when the assumptions for the function in the population level also hold for each random realization almost surely (e.g., Lipschitzness of each realization of the stochastic gradient). To address this issue, we propose a new Accelerated Bilevel Optimization algorithm named AccBO. The algorithm updates the upper-level variable by normalized stochastic gradient descent with recursive momentum and the lower-level variable by the stochastic Nesterov accelerated gradient descent algorithm with averaging. We prove that our algorithm achieves an oracle complexity of $\widetilde{O}(\epsilon^{-3})$ to find an $\epsilon$-stationary point, when the lower-level stochastic gradient has a small variance $O(\epsilon)$. Our proof relies on a novel lemma characterizing the dynamics of stochastic Nesterov accelerated gradient descent algorithm under distribution drift with high probability for the lower-level variable, which is of independent interest and also plays a crucial role in analyzing the hypergradient estimation error over time. Experimental results on various tasks confirm that our proposed algorithm achieves the predicted theoretical acceleration and significantly outperforms baselines in bilevel optimization.

Computer Vision · Classification

Yitian Zhang, n n, Xu Ma, Huan Wang, Ke Ma, Stephen Chen, Derek Hu, Yun Fu

Vision Transformers (ViT) is known for its scalability. In this work, we target to scale down a ViT to fit in an environment with dynamic-changing resource constraints. We observe that smaller ViTs are intrinsically the sub-networks of a larger ViT with different widths. Thus, we propose a general framework, named Scala, to enable a single network to represent multiple smaller ViTs with flexible inference capability, which aligns with the inherent design of ViT to vary from widths. Concretely, Scala activates several subnets during training, introduces Isolated Activation to disentangle the smallest sub-network from other subnets, and leverages Scale Coordination to ensure each sub-network receives simplified, steady, and accurate learning objectives. Comprehensive empirical validations on different tasks demonstrate that with only one-shot training, Scala learns slimmable representation without modifying the original ViT structure and matches the performance of Separate Training. Compared with the prior art, Scala achieves an average improvement of 1.6% on ImageNet-1K with fewer parameters.

Shayan Kiyani, George J. Pappas, Hamed Hassani

Conditional validity and length efficiency are two crucial aspects of conformal prediction (CP). Conditional validity ensures accurate uncertainty quantification for data subpopulations, while proper length efficiency ensures that the prediction sets remain informative. Despite significant efforts to address each of these issues individually, a principled framework that reconciles these two objectives has been missing in the CP literature. In this paper, we develop Conformal Prediction with Length-Optimization (CPL) - a novel and practical framework that constructs prediction sets with (near-) optimal length while ensuring conditional validity under various classes of covariate shifts, including the key cases of marginal and group-conditional coverage. In the infinite sample regime, we provide strong duality results which indicate that CPL achieves conditional validity and length optimality. In the finite sample regime, we show that CPL constructs conditionally valid prediction sets. Our extensive empirical evaluations demonstrate the superior prediction set size performance of CPL compared to state-of-the-art methods across diverse real-world and synthetic datasets in classification, regression, and large language model-based multiple choice question answering. An Implementation of our algorithm can be accessed at the following link: https://github.com/shayankiyani98/CP.

Ruiqi Zhong, Heng Wang, Dan Klein, Jacob Steinhardt

To make sense of massive data, we often first fit simplified models and then interpret the parameters; for example, we cluster the text embeddings and then interpret the mean parameters of each cluster.However, these parameters are often high-dimensional and hard to interpret.To make model parameters directly interpretable, we introduce a family of statistical models---including clustering, time series, and classification models---parameterized by *natural language predicates*. For example, a cluster of text about COVID could be parameterized by the predicate ``*discusses COVID*''.To learn these statistical models effectively, we develop a model-agnostic algorithm that optimizes continuous relaxations of predicate parameters with gradient descent and discretizes them by prompting language models (LMs).Finally, we apply our framework to a wide range of problems: taxonomizing user chat dialogues, characterizing how they evolve across time, finding categories where one language model is better than the other, clustering math problems based on subareas, and explaining visual features in memorable images.Our framework is highly versatile, applicable to both textual and visual domains, can be easily steered to focus on specific properties (e.g. subareas), and explains sophisticated concepts that classical methods (e.g. n-gram analysis) struggle to produce.

Computer Vision

Mengyuan Chen, Junyu Gao, Changsheng Xu

A straightforward pipeline for zero-shot out-of-distribution (OOD) detection involves selecting potential OOD labels from an extensive semantic pool and then leveraging a pre-trained vision-language model to perform classification on both in-distribution (ID) and OOD labels. In this paper, we theorize that enhancing performance requires expanding the semantic pool, while increasing the expected probability of selected OOD labels being activated by OOD samples, and ensuring low mutual dependence among the activations of these OOD labels. A natural expansion manner is to adopt a larger lexicon; however, the inevitable introduction of numerous synonyms and uncommon words fails to meet the above requirements, indicating that viable expansion manners move beyond merely selecting words from a lexicon. Since OOD detection aims to correctly classify input images into ID/OOD class groups, we can "make up" OOD label candidates which are not standard class names but beneficial for the process. Observing that the original semantic pool is comprised of unmodified specific class names, we correspondingly construct a conjugated semantic pool (CSP) consisting of modified superclass names, each serving as a cluster center for samples sharing similar properties across different categories. Consistent with our established theory, expanding OOD label candidates with the CSP satisfies the requirements and outperforms existing works by 7.89% in FPR95. Codes are available in https://github.com/MengyuanChen21/NeurIPS2024-CSP.

Interpretability

Roland S. Zimmermann, David Klindt, Wieland Brendel

In today’s era, whatever we can measure at scale, we can optimize. So far, measuring the interpretability of units in deep neural networks (DNNs) for computer vision still requires direct human evaluation and is not scalable. As a result, the inner workings of DNNs remain a mystery despite the remarkable progress we have seen in their applications. In this work, we introduce the first scalable method to measure the per-unit interpretability in vision DNNs. This method does not require any human evaluations, yet its prediction correlates well with existing human interpretability measurements. We validate its predictive power through an interventional human psychophysics study. We demonstrate the usefulness of this measure by performing previously infeasible experiments: (1) A large-scale interpretability analysis across more than 70 million units from 835 computer vision models, and (2) an extensive analysis of how units transform during training. We find an anticorrelation between a model's downstream classification performance and per-unit interpretability, which is also observable during model training. Furthermore, we see that a layer's location and width influence its interpretability.

Zhenyu Wang, Ya-Li Li, Hengshuang Zhao, Shengjin Wang

The current trend in computer vision is to utilize one universal model to address all various tasks. Achieving such a universal model inevitably requires incorporating multi-domain data for joint training to learn across multiple problem scenarios. In point cloud based 3D object detection, however, such multi-domain joint training is highly challenging, because large domain gaps among point clouds from different datasets lead to the severe domain-interference problem. In this paper, we propose OneDet3D, a universal one-for-all model that addresses 3D detection across different domains, including diverse indoor and outdoor scenes, within the same framework and only one set of parameters. We propose the domain-aware partitioning in scatter and context, guided by a routing mechanism, to address the data interference issue, and further incorporate the text modality for a language-guided classification to unify the multi-dataset label spaces and mitigate the category interference issue. The fully sparse structure and anchor-free head further accommodate point clouds with significant scale disparities. Extensive experiments demonstrate the strong universal ability of OneDet3D to utilize only one trained model for addressing almost all 3D object detection tasks (Fig. 1). We will open-source the code for future research and applications.

Computer Vision · Classification

Yuhan Zhu, Yuyang Ji, Zhiyu Zhao, Gangshan Wu, Limin Wang

Pre-trained vision-language models (VLMs) have shown impressive results in various visual classification tasks.However, we often fail to fully unleash their potential when adapting them for new concept understanding due to limited information on new classes.To address this limitation, we introduce a novel adaptation framework, AWT (Augment, Weight, then Transport). AWT comprises three key components: augmenting inputs with diverse visual perspectives and enriched class descriptions through image transformations and language models; dynamically weighting inputs based on the prediction entropy; and employing optimal transport to mine semantic correlations in the vision-language space.AWT can be seamlessly integrated into various VLMs, enhancing their zero-shot capabilities without additional training and facilitating few-shot learning through an integrated multimodal adapter module.We verify AWT in multiple challenging scenarios, including zero-shot and few-shot image classification, zero-shot video action recognition, and out-of-distribution generalization. AWT consistently outperforms the state-of-the-art methods in each setting. In addition, our extensive studies further demonstrate AWT's effectiveness and adaptability across different VLMs, architectures, and scales.

Hongren Yan, Yuhua Qian, Furong Peng, Jiachen Luo, zheqing zhu, Feijiang Li

Neural Collapse (NC) was a recently discovered phenomenon that the output features and the classifier weights of the neural network converge to optimal geometric structures at the Terminal Phase of Training (TPT) under various losses. However, the relationship between these optimal structures at TPT and the classification performance remains elusive, especially in imbalanced learning. Even though it is noticed that fixing the classifier to an optimal structure can mitigate the minority collapse problem, the performance is still not comparable to the classical imbalanced learning methods with a learnable classifier. In this work, we find that the optimal structure can be designed to represent a better classification rule, and thus achieve better performance. In particular, we justify that, to achieve better classification, the features from the minor classes should align with more directions. This justification then yields a decision rule called the Generalized Classification Rule (GCR) and we also term these directions as the centers of the classes. Then we study the NC under an MSE-type loss via the Unconstrained Features Model (UFM) framework where (1) the features from a class tend to collapse to the mean of the corresponding centers of that class (named Neural Collapse to Multiple Centers (NCMC)) at the global optimum, and (2) the original classifier approximates a surrogate to GCR when NCMC occurs. Based on the analysis, we develop a strategy for determining the number of centers and propose a Cosine Loss function for the fixed classifier that induces NCMC. Our experiments have shown that the Cosine Loss can induce NCMC and has performance on long-tail classification comparable to the classical imbalanced learning methods.

Deep Learning · Everything Else

Francesco D'Angelo, Maksym Andriushchenko, Aditya Vardhan Varre, Nicolas Flammarion

Weight decay is a broadly used technique for training state-of-the-art deep networks from image classification to large language models. Despite its widespread usage and being extensively studied in the classical literature, its role remains poorly understood for deep learning. In this work, we highlight that the role of weight decay in modern deep learning is different from its regularization effect studied in classical learning theory. For deep networks on vision tasks trained with multipass SGD, we show how weight decay modifies the optimization dynamics enhancing the ever-present implicit regularization of SGD via the *loss stabilization mechanism*. In contrast, for large language models trained with nearly one-epoch training, we describe how weight decay balances the *bias-variance tradeoff* in stochastic optimization leading to lower training loss and improved training stability. Overall, we present a unifying perspective from ResNets on vision tasks to LLMs: weight decay is never useful as an explicit regularizer but instead changes the training dynamics in a desirable way.

Graph Neural Networks

ZEYU ZHANG, Lu Li, Shuyan Wan, Wang, Zhiyi Wang, Zhiyuan Lu, Dong Hao, Wanli Li

Signed graphs can model friendly or antagonistic relations where edges are annotated with a positive or negative sign. The main downstream task in signed graph analysis is $\textit{link sign prediction}$. Signed Graph Neural Networks (SGNNs) have been widely used for signed graph representation learning. While significant progress has been made in SGNNs research, two issues (i.e., graph sparsity and unbalanced triangles) persist in the current SGNN models. We aim to alleviate these issues through data augmentation ($\textit{DA}$) techniques which have demonstrated effectiveness in improving the performance of graph neural networks. However, most graph augmentation methods are primarily aimed at graph-level and node-level tasks (e.g., graph classification and node classification) and cannot be directly applied to signed graphs due to the lack of side information (e.g., node features and label information) in available real-world signed graph datasets. Random $\textit{DropEdge} $is one of the few $\textit{DA}$ methods that can be directly used for signed graph data augmentation, but its effectiveness is still unknown. In this paper, we first provide the generalization bound for the SGNN model and demonstrate from both experimental and theoretical perspectives that the random $\textit{DropEdge}$ cannot improve the performance of link sign prediction. Therefore, we propose a novel signed graph augmentation method, $\underline{S}$igned $\underline{G}$raph $\underline{A}$ugmentation framework (SGA). Specifically, SGA first integrates a structure augmentation module to detect candidate edges solely based on network information. Furthermore, SGA incorporates a novel strategy to select beneficial candidates. Finally, SGA introduces a novel data augmentation perspective to enhance the training process of SGNNs. Experiment results on six real-world datasets demonstrate that SGA effectively boosts the performance of diverse SGNN models, achieving improvements of up to 32.3\% in F1-micro for SGCN on the Slashdot dataset in the link sign prediction task.

Multimodal Models

Chao Yi, Yuhang He, De-Chuan Zhan, Han-Jia Ye

Vision Language Models (VLMs) excel in zero-shot image classification by pairing images with textual category names. The expanding variety of Pre-Trained VLMs enhances the likelihood of identifying a suitable VLM for specific tasks. To better reuse the VLM resource and fully leverage its potential on different zero-shot image classification tasks, a promising strategy is selecting appropriate Pre-Trained VLMs from the VLM Zoo, relying solely on the text data of the target dataset without access to the dataset’s images. In this paper, we analyze two inherent challenges in assessing the ability of a VLM in this Language-Only VLM selection: the “Modality Gap”—the disparity in VLM’s embeddings across two different modalities, making text a less reliable substitute for images; and the “Capability Gap”— the discrepancy between the VLM’s overall ranking and its ranking for target dataset, hindering direct prediction of a model’s dataset-specific performance from its general performance. We propose VLM Selection With gAp Bridging (SWAB) to mitigate the negative impact of two gaps. SWAB first adopts optimal transport to capture the relevance between open-source and target datasets with a transportation matrix. It then uses this matrix to transfer useful statistics of VLMs from open-source datasets to the target dataset for bridging two gaps. By bridging two gaps to obtain better substitutes for test images, SWAB can accurately predict the performance ranking of different VLMs on the target task without the need for the dataset’s images. Experiments across various VLMs and image classification datasets validate SWAB’s effectiveness. Code is available at: https://github.com/YCaigogogo/SWAB.

Trustworthy Machine Learning

Qizhang Li, Yiwen Guo, Wangmeng Zuo, Hao Chen

Adversarial prompts (or say, adversarial examples) generated using gradient-based methods exhibit outstanding performance in performing automatic jailbreak attacks against safety-aligned LLMs. Nevertheless, due to the discrete nature of texts, the input gradient of LLMs struggles to precisely reflect the magnitude of loss change that results from token replacements in the prompt, leading to limited attack success rates against safety-aligned LLMs, even in the *white-box* setting. In this paper, we explore a new perspective on this problem, suggesting that it can be alleviated by leveraging innovations inspired in transfer-based attacks that were originally proposed for attacking *black-box* image classification models. For the first time, we appropriate the ideologies of effective methods among these transfer-based attacks, *i.e.*, Skip Gradient Method and Intermediate Level Attack, into gradient-based adversarial prompt generation and achieve significant performance gains without introducing obvious computational cost. Meanwhile, by discussing mechanisms behind the gains, new insights are drawn, and proper combinations of these methods are also developed. Our empirical results show that 87% of the query-specific adversarial suffixes generated by the developed combination can induce Llama-2-7B-Chat to produce the output that exactly matches the target string on AdvBench. This match rate is 33% higher than that of a very strong baseline known as GCG, demonstrating advanced discrete optimization for adversarial prompt generation against LLMs. In addition, without introducing obvious cost, the combination achieves >30% absolute increase in attack success rates compared with GCG when generating both query-specific (38% ->68%) and universal adversarial prompts (26.68% -> 60.32%) for attacking the Llama-2-7B-Chat model on AdvBench.Code at: https://github.com/qizhangli/Gradient-based-Jailbreak-Attacks.

Deep Learning · Algorithms

Sike Wang, Pan Zhou, Jia Li, Hua Huang

Second-order optimizers, maintaining a matrix termed a preconditioner, are superior to first-order optimizers in both theory and practice.The states forming the preconditioner and its inverse root restrict the maximum size of models trained by second-order optimizers. To address this, compressing 32-bit optimizer states to lower bitwidths has shown promise in reducing memory usage. However, current approaches only pertain to first-order optimizers. In this paper, we propose the first 4-bit second-order optimizers, exemplified by 4-bit Shampoo, maintaining performance similar to that of 32-bit ones. We show that quantizing the eigenvector matrix of the preconditioner in 4-bit Shampoo is remarkably better than quantizing the preconditioner itself both theoretically and experimentally. By rectifying the orthogonality of the quantized eigenvector matrix, we enhance the approximation of the preconditioner's eigenvector matrix, which also benefits the computation of its inverse 4-th root. Besides, we find that linear square quantization slightly outperforms dynamic tree quantization when quantizing second-order optimizer states. Evaluation on various networks for image classification and natural language modeling demonstrates that our 4-bit Shampoo achieves comparable performance to its 32-bit counterpart while being more memory-efficient.