Author name cluster

Han Zhao

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

55 papers

1 author row

TMLR Journal 2026 Journal Article

Adaptive Online Mirror Descent for Tchebycheff Scalarization in Multi-Objective Learning

Meitong Liu
Xiaoyuan Zhang
Chulin Xie
Kate Donahue
Han Zhao

Multi-objective learning (MOL) aims to learn under multiple potentially conflicting objectives and strike a proper balance. While recent preference-guided MOL methods often rely on additional optimization objectives or constraints, we consider the classic Tchebycheff scalarization (TCH) that naturally allows for locating solutions with user-specified trade-offs. Due to its minimax formulation, directly optimizing TCH often leads to training oscillation and stagnation. In light of this limitation, we propose an adaptive online mirror descent algorithm for TCH, called (Ada)OMD-TCH. One of our main ingredients is an adaptive online-to-batch conversion that significantly improves solution optimality over traditional conversion in practice while maintaining the same theoretical convergence guarantees. We show that (Ada)OMD-TCH achieves a convergence rate of $\mathcal O(\sqrt{\log m/T})$, where $m$ is the number of objectives and $T$ is the number of rounds, providing a tighter dependency on $m$ in the offline setting compared to existing work. Empirically, we demonstrate on both synthetic problems and federated learning tasks that (Ada)OMD-TCH effectively smooths the training process and yields preference-guided, specific, diverse, and fair solutions.

TIST Journal 2026 Journal Article

An Improved Autoregressive Evaluation Paradigm for Large Language Models

Jipeng Zhang
Rui Pan
Yuzheng Hu
KaShun Shum
Guanyu Yao
Xiang Liu
Renjie Pi
Hanze Dong

The AI community has witnessed the emergence of various chat-style Large Language Models (LLMs) since the advent of ChatGPT. Despite significant progress in this area, evaluating these models remains a substantial challenge. The evaluations provided by humans or GPT-4 oracles are often taken as the gold standard, but they are neither automatic nor scalable. More recently, a series of (open source) LLM-based judge models have been introduced, yet they often exhibit model-specific biases, e.g., a LLaMA-family judge favors a LLaMA-family model. On the other hand, autoregressive evaluation metrics, which holds the potential to address the aforementioned issues, remains underexplored. Among them, likelihood-based metrics such as perplexity and Negative Log-Likelihood (NLL) are widely adopted and has proven effective in tracking the pre-training progress of LLMs. However, they struggle to evaluate the generation capabilities of fine-tuned models due to exposure bias, a phenomenon where the distribution of the model’s output gradually deviates from the ground-truth during inference. To address this key issue, in this article, we propose a novel autoregressive metric, Normalized Discounted Cumulative Gain (NDCG), to improve the evaluation of fine-tuned LLMs. Our experimental results demonstrate that NDCG significantly outperforms likelihood-based metrics: it shows over 45% improvement in both Spearman and Kendall’s tau correlation coefficients for commonsense QA tasks, and aligns more closely with GPT-4 Elo rankings for instruction-tuned models.

AAAI Conference 2026 Conference Paper

ReconVLA: Reconstructive Vision-Language-Action Model as Effective Robot Perceiver

Wenxuan Song
Ziyang Zhou
Han Zhao
Jiayi Chen
Pengxiang Ding
Haodong Yan
Yuxin Huang
Feilong Tang

Recent advances in Vision-Language-Action (VLA) models have enabled robotic agents to integrate multimodal understanding with action execution. However, our empirical analysis reveals that current VLAs struggle to allocate visual attention to target regions. Instead, visual attention is always dispersed. To guide the visual attention grounding on the correct target, we propose ReconVLA, a reconstructive VLA model with an implicit grounding paradigm. Conditioned on the model's visual outputs, a diffusion transformer aims to reconstruct the gaze region of the image, which corresponds to the target manipulated objects. This process prompts the VLA model to learn fine-grained representations and accurately allocate visual attention, thus effectively leveraging task-specific visual information and conducting precise manipulation. Moreover, we curate a large-scale pretraining dataset comprising over 100k trajectories and 2 million data samples from open-source robotic datasets, further boosting the model’s generalization in visual reconstruction. Extensive experiments in simulation and the real world demonstrate the superiority of our implicit grounding method, showcasing its capabilities of precise manipulation and generalization.

PDF Details DOI

AAAI Conference 2026 Conference Paper

VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Model

Yihao Wang
Pengxiang Ding
Lingxiao Li
Can Cui
Zirui Ge
Xinyang Tong
Wenxuan Song
Han Zhao

Vision-Language-Action (VLA) models typically bridge the gap between perceptual and action spaces by pre-training a large-scale Vision-Language Model (VLM) on robotic data. While this approach greatly enhances performance, it also incurs significant training costs. In this paper, we investigate how to effectively bridge vision-language (VL) representations to action (A). We introduce VLA-Adapter, a novel paradigm designed to reduce the reliance of VLA models on large-scale VLMs and extensive pre-training. To this end, we first systematically analyze the effectiveness of various VL conditions and present key findings on which conditions are essential for bridging perception and action spaces. Based on these insights, we propose a lightweight Policy module with Bridge Attention, which autonomously injects the optimal condition into the action space. In this way, our method achieves high performance using only a 0.5B-parameter backbone, without any robotic data pre-training. Extensive experiments on both simulated and real-world robotic benchmarks show that VLA-Adapter not only achieves state-of-the-art level performance, but also offers the fast inference speed reported to date. Furthermore, thanks to the proposed advanced bridging paradigm, VLA-Adapter enables the training of a powerful VLA model on a single consumer-grade GPU, greatly lowering the barrier to deploying VLA model.

PDF Details DOI

TMLR Journal 2026 Journal Article

Weakly-Supervised Disentangled Representation Learning via Filter-Based Adaptive Swapping

Zhenyu Zong
Qidi Wang
Simon Yu
Hongpeng Cao
Yanbing Mao
Han Zhao
Lui Sha
Huajie Shao

Disentangled representation learning (DRL) aims to uncover semantically meaningful latent factors from observed data, thereby improving both interpretability and generalization of machine learning (ML) models. Despite remarkable progress, unsupervised DRL cannot achieve complete disentanglement without inductive biases or supervision. To address this challenge, existing approaches either rely on full supervision, which demands extensive manual labeling, or weak supervision, which involves complex training strategies that often result in unstable training. To address these limitations, we propose Filter-VAE, a weakly supervised variational autoencoder (VAE) that introduces a filter-based adaptive swapping strategy to learn stable and meaningful disentangled representations. Specifically, a relevance filter removes semantically meaningless latent factors, while an adaptive swapping filter exchanges those latent factors that have reached stability. With these two filters, Filter-VAE adaptively swaps only stable and semantically aligned latent factors, leading to robust and meaningful representations. We evaluate Filter-VAE on three standard benchmarks and our created traffic sign dataset in two downstream tasks: disentanglement and adversarial robustness. Experimental results demonstrate that Filter-VAE achieves strong disentanglement performance with reduced supervision and delivers remarkable robustness against diverse adversarial attacks and corruptions. The code is released at https://github.com/ZY-Zong/Filter-VAE.git.

NeurIPS Conference 2025 Conference Paper

A Snapshot of Influence: A Local Data Attribution Framework for Online Reinforcement Learning

Yuzheng Hu
Fan Wu
Haotian Ye
David Forsyth
James Zou
Nan Jiang
Jiaqi Ma
Han Zhao

Online reinforcement learning (RL) excels in complex, safety-critical domains but suffers from sample inefficiency, training instability, and limited interpretability. Data attribution provides a principled way to trace model behavior back to training samples, yet existing methods assume fixed datasets, which is violated in online RL where each experience both updates the policy and shapes future data collection. In this paper, we initiate the study of data attribution for online RL, focusing on the widely used Proximal Policy Optimization (PPO) algorithm. We start by establishing a local attribution framework, interpreting model checkpoints with respect to the records in the recent training buffer. We design two target functions, capturing agent action and cumulative return respectively, and measure each record's contribution through gradient similarity between its training loss and these targets. We demonstrate the power of this framework through three concrete applications: diagnosis of learning, temporal analysis of behavior formation, and targeted intervention during training. Leveraging this framework, we further propose an algorithm, iterative influence-based filtering (IIF), for online RL training that iteratively performs experience filtering to refine policy updates. Across standard RL benchmarks (classic control, navigation, locomotion) to RLHF for large language models, IIF reduces sample complexity, speeds up training, and achieves higher returns. Together, these results open a new direction for making online RL more interpretable, efficient, and effective.

AAAI Conference 2025 Conference Paper

Cobra: Extending Mamba to Multi-Modal Large Language Model for Efficient Inference

Han Zhao
Min Zhang
Wei Zhao
Pengxiang Ding
Siteng Huang
Donglin Wang

In recent years, applying multi-modal large language models (MLLMs) in various fields has achieved remarkable success. However, as the foundation model for many downstream tasks, MLLMs comprise the well-known Transformer network, which has a less efficient quadratic computation complexity. In this study, we introduce Cobra, a multi-modal large-scale language model built upon a state-space model, which has demonstrated significant potential in efficiently handling long sequences with fast inference and linear scalability concerning sequence length. Specifically, Cobra involves replacing Transformer-based backbone models (e.g., LLaMA or Phi) with pre-trained Mamba language models. We then empirically explore effective strategies for aligning visual and textual modalities and integrating various pre-trained Mamba model variants with visual encoders. Experiments across various multi-modal benchmarks demonstrate that: (i) Cobra performs 3× ∼ 4× faster than the most computationally efficient state-of-the-art methods, e.g., LLaVA-Phi and MobileVLM v2. Additionally, its performance is significantly enhanced thanks to the implementation of linear sequential modeling. (ii) Cobra fine-tunes a small parameter (∼48% of model parameters), leading to a significant improvement in overall performance compared to LLaVA.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Efficient Utility-Preserving Machine Unlearning with Implicit Gradient Surgery

Shiji Zhou
Tianbai Yu
Zhi Zhang
Heng Chang
Xiao Zhou
Dong Wu
Han Zhao

Machine unlearning (MU) aims to efficiently remove sensitive or harmful memory from a pre-trained model. The key challenge is to balance the potential tradeoff between unlearning efficacy and utility preservation, which involves forgetting undesirable information as defined while maintaining the model's original performance. One potential way to tackle this problem is to use multi-objective optimization to jointly optimize both the unlearning and utility preservation objectives. However, existing multi-objective methods only guarantee finding a Pareto-optimal solution without fine-grained control, which causes under-optimization of the unlearning objective. To this end, we first model MU as a constrained optimization problem, that is, optimizing the unlearning objective under the constraint of a bounded increase for utility loss. We then show that solving this optimization problem is equivalent to unilateral gradient surgery on the unlearning objective. To resolve the additional computational cost brought by gradient surgery, we propose an implicit gradient surgery method, which approximates the solution to the aforementioned constrained optimization problem via only one backpropagation, thereby achieving efficient utility-preserving MU. Theoretically, we provide a tight convergence analysis of the algorithm. Empirically, our extensive experiments show that the proposed algorithm achieves better tradeoff results than existing baselines. Codes are available at https: //github. com/anseryuer/EUPMU-Efficient-Utility-Preserving-Machine-Unlearning.

NeurIPS Conference 2025 Conference Paper

GraSS: Scalable Data Attribution with Gradient Sparsification and Sparse Projection

Pingbang Hu
Joseph Melkonian
Weijing Tang
Han Zhao
Jiaqi Ma

Gradient-based data attribution methods, such as influence functions, are critical for understanding the impact of individual training samples without requiring repeated model retraining. However, their scalability is often limited by the high computational and memory costs associated with per-sample gradient computation. In this work, we propose GraSS, a novel gradient compression algorithm and its variants FactGraSS for linear layers specifically, that explicitly leverage the inherent sparsity of per-sample gradients to achieve sub-linear space and time complexity. Extensive experiments demonstrate the effectiveness of our approach, achieving substantial speedups while preserving data influence fidelity. In particular, FactGraSS achieves up to 165% faster throughput on billion-scale models compared to the previous state-of-the-art baselines. Our code is publicly available at https: //github. com/TRAIS-Lab/GraSS.

TMLR Journal 2025 Journal Article

Localize-and-Stitch: Efficient Model Merging via Sparse Task Arithmetic

Yifei He
Yuzheng Hu
Yong Lin
Tong Zhang
Han Zhao

Model merging offers an effective strategy to combine the strengths of multiple finetuned models into a unified model that preserves the specialized capabilities of each. Existing methods merge models in a global manner, performing arithmetic operations across all model parameters. However, such global merging often leads to task interference, degrading the performance of the merged model. In this work, we introduce Localize-and-Stitch, a novel approach that merges models in a localized way. Our algorithm works in two steps: i) Localization: identify tiny ($1\%$ of the total parameters) localized regions in the finetuned models containing essential skills for the downstream tasks, and ii) Stitching: reintegrate only these essential regions back into the pretrained model for task synergy. We demonstrate that our approach effectively locates sparse regions responsible for finetuned performance, and the localized regions could be treated as compact and interpretable representations of the finetuned models (tasks). Empirically, we evaluate our method on various vision and language benchmarks, showing that it outperforms existing model merging methods under different data availability scenarios. Beyond strong empirical performance, our algorithm also facilitates model compression and preserves pretrained knowledge, enabling flexible and continual skill composition from multiple finetuned models with minimal storage and computational overhead.

NeurIPS Conference 2025 Conference Paper

MergeBench: A Benchmark for Merging Domain-Specialized LLMs

Yifei He
Siqi Zeng
Yuzheng Hu
Rui Yang
Tong Zhang
Han Zhao

Model merging provides a scalable alternative to multi-task training by combining specialized finetuned models through parameter arithmetic, enabling efficient deployment without the need for joint training or access to all task data. While recent methods have shown promise, existing evaluations are limited in both model scale and task diversity, leaving open questions about their applicability to large, domain-specialized LLMs. To tackle the challenges, we introduce MergeBench, a comprehensive evaluation suite designed to assess model merging at scale. MergeBench builds on state-of-the-art open-source language models, including Llama and Gemma families at 2B to 9B scales, and covers five key domains: instruction following, mathematics, multilingual understanding, coding and safety. We standardize finetuning and evaluation protocols, and assess eight representative merging methods across multi-task performance, forgetting and runtime efficiency. Based on extensive experiments, we provide practical guidelines for algorithm selection and share insights showing that model merging tends to perform better on stronger base models, with techniques such as merging coefficient tuning and sparsification improving knowledge retention. However, several challenges remain, including the computational cost on large models, the gap for in-domain performance compared to multi-task models, and the underexplored role of model merging in standard LLM training pipelines. We hope MergeBench provides a foundation for future research to advance the understanding and practical application of model merging.

NeurIPS Conference 2025 Conference Paper

SSR: Enhancing Depth Perception in Vision-Language Models via Rationale-Guided Spatial Reasoning

Yang Liu
Ming Ma
Xiaomin Yu
Pengxiang Ding
Han Zhao
Mingyang Sun
Siteng Huang
Donglin Wang

Despite impressive advancements in Visual-Language Models (VLMs) for multi-modal tasks, their reliance on RGB inputs limits precise spatial understanding. Existing methods for integrating spatial cues, such as point clouds or depth, either require specialized sensors or fail to effectively exploit depth information for higher-order reasoning. To this end, we propose a novel Spatial Sense and Reasoning method, dubbed SSR, a novel framework that transforms raw depth data into structured, interpretable textual rationales. These textual rationales serve as meaningful intermediate representations to significantly enhance spatial reasoning capabilities. Additionally, we leverage knowledge distillation to compress the generated rationales into compact latent embeddings, which facilitate resource-efficient and plug-and-play integration into existing VLMs without retraining. To enable comprehensive evaluation, we introduce a new dataset named SSR-CoT, a million-scale visual-language reasoning dataset enriched with intermediate spatial reasoning annotations, and present SSRBench, a comprehensive multi-task benchmark. Extensive experiments on multiple benchmarks demonstrate SSR substantially improves depth utilization and enhances spatial reasoning, thereby advancing VLMs toward more human-like multi-modal understanding. Project page: https: //yliu-cs. github. io/SSR.

NeurIPS Conference 2025 Conference Paper

Taming Hyperparameter Sensitivity in Data Attribution: Practical Selection Without Costly Retraining

Weiyi Wang
Junwei Deng
Yuzheng Hu
Shiyuan Zhang
Xirui Jiang
Runting Zhang
Han Zhao
Jiaqi Ma

Data attribution methods, which quantify the influence of individual training data points on a machine learning model, have gained increasing popularity in data-centric applications in modern AI. Despite a recent surge of new methods developed in this space, the impact of hyperparameter tuning in these methods remains under-explored. In this work, we present the first large-scale empirical study to understand the hyperparameter sensitivity of common data attribution methods. Our results show that most methods are indeed sensitive to certain key hyperparameters. However, unlike typical machine learning algorithms---whose hyperparameters can be tuned using computationally-cheap validation metrics---evaluating data attribution performance often requires retraining models on subsets of training data, making such metrics prohibitively costly for hyperparameter tuning. This poses a critical open challenge for the practical application of data attribution methods. To address this challenge, we advocate for better theoretical understandings of hyperparameter behavior to inform efficient tuning strategies. As a case study, we provide a theoretical analysis of the regularization term that is critical in many variants of influence function methods. Building on this analysis, we propose a lightweight procedure for selecting the regularization value without model retraining, and validate its effectiveness across a range of standard data attribution benchmarks. Overall, our study identifies a fundamental yet overlooked challenge in the practical application of data attribution, and highlights the importance of careful discussion on hyperparameter selection in future method development.

NeurIPS Conference 2025 Conference Paper

Understanding and Improving Adversarial Robustness of Neural Probabilistic Circuits

Weixin Chen
Han Zhao

Neural Probabilistic Circuits (NPCs), a new class of concept bottleneck models, comprise an attribute recognition model and a probabilistic circuit for reasoning. By integrating the outputs from these two modules, NPCs produce compositional and interpretable predictions. While offering enhanced interpretability and high performance on downstream tasks, the neural-network-based attribute recognition model remains a black box. This vulnerability allows adversarial attacks to manipulate attribute predictions by introducing carefully crafted subtle perturbations to input images, potentially compromising the final predictions. In this paper, we theoretically analyze the adversarial robustness of NPC and demonstrate that it only depends on the robustness of the attribute recognition model and is independent of the robustness of the probabilistic circuit. Moreover, we propose RNPC, the first robust neural probabilistic circuit against adversarial attacks on the recognition module. RNPC introduces a novel class-wise integration for inference, ensuring a robust combination of outputs from the two modules. Our theoretical analysis demonstrates that RNPC exhibits provably improved adversarial robustness compared to NPC. Empirical results on image classification tasks show that RNPC achieves superior adversarial robustness compared to existing concept bottleneck models while maintaining high accuracy on benign inputs. The code is available at https: //github. com/uiuctml/RNPC.

TMLR Journal 2025 Journal Article

Understanding Emergent In-Context Learning from a Kernel Regression Perspective

Chi Han
Ziqi Wang
Han Zhao
Heng Ji

Large language models (LLMs) have initiated a paradigm shift in transfer learning. In contrast to the classic pretraining-then-finetuning procedure, in order to use LLMs for downstream prediction tasks, one only needs to provide a few demonstrations, known as in-context examples, without adding more or updating existing model parameters. This in-context learning (ICL) capability of LLMs is intriguing, and it is not yet fully understood how pretrained LLMs acquire such capabilities. In this paper, we investigate the reason why a transformer-based language model can accomplish in-context learning after pre-training on a general language corpus by proposing a kernel-regression perspective of understanding LLMs' ICL behaviors when faced with in-context examples. More concretely, we first prove that Bayesian inference on in-context prompts can be asymptotically understood as kernel regression $\hat y = \sum_i y_i K(x, x_i)/\sum_i K(x, x_i)$ as the number of in-context demonstrations grows. Then, we empirically investigate the in-context behaviors of language models. We find that during ICL, the attention and hidden features in LLMs match the behaviors of a kernel regression. Finally, our theory provides insights into multiple phenomena observed in the ICL field: why retrieving demonstrative samples similar to test samples can help, why ICL performance is sensitive to the output formats, and why ICL accuracy benefits from selecting in-distribution and representative samples.

TMLR Journal 2024 Journal Article

A General-Purpose Multi-Modal OOD Detection Framework

Viet Quoc Duong
Qiong Wu
Zhengyi Zhou
Eric Zavesky
WenLing Hsu
Han Zhao
Huajie Shao

Out-of-distribution (OOD) detection seeks to identify test samples that deviate from the training data, which is critical to ensuring the safety and reliability of machine learning (ML) systems. While a plethora of methods have been developed to detect uni-modal OOD samples, only a few have focused on multi-modal OOD detection. Current contrastive learning-based methods primarily address multi-modal OOD detection in a scenario where an image is not related to the class labels in training data. However, ML systems in the real-world applications may encounter a broader spectrum of anomalies caused by different factors like systematic errors in labeling, environmental changes, and sensor malfunctions. Hence, we propose a new method to be able to simultaneously detect anomalies from multiple different OOD scenarios, arising from fine-grained image features and textual descriptions, instead of large categorical information. To achieve this goal, we propose a general-purpose weakly-supervised OOD detection framework, called WOOD, that combines a binary classifier and a contrastive learning module to reap the benefits of both. In order to better distinguish in-distribution (ID) samples from OOD ones, we employ the Hinge loss to constrain the similarity of their latent representations. Moreover, we devise a new scoring metric that fuses predictions from both the binary classifier and contrastive learning to enhance OOD detection. Extensive experimental results on multiple benchmarks demonstrate that the proposed WOOD significantly outperforms the state-of-the-art methods for multi-modal OOD detection. Importantly, our approach can achieve superior detection performance in a variety of OOD scenarios.

YNIMG Journal 2024 Journal Article

Characterizing the role of the microbiota-gut-brain axis in cerebral small vessel disease: An integrative multi‑omics study

Yu Song
Xia Zhou
Han Zhao
Wenming Zhao
Zhongwu Sun
Jiajia Zhu
Yongqiang Yu

BACKGROUND: Prior efforts have revealed changes in gut microbiome, circulating metabolome, and multimodal neuroimaging features in cerebral small vessel disease (CSVD). However, there is a paucity of research integrating the multi-omic information to characterize the role of the microbiota-gut-brain axis in CSVD. METHODS: We collected gut microbiome, fecal and blood metabolome, multimodal magnetic resonance imaging data from 37 CSVD patients with white matter hyperintensities and 46 healthy controls. Between-group comparison was performed to identify the differential gut microbial taxa, followed by performance of multi-stage microbiome-metabolome-neuroimaging-neuropsychology correlation analyses in CSVD patients. RESULTS: Our data showed both depleted and enriched gut microbes in CSVD patients. Among the differential microbes, Haemophilus and Akkermansia were associated with a range of metabolites enriched for Aminoacyl-tRNA biosynthesis pathway. Furthermore, the affected metabolites were associated with neuroimaging measures involving gray matter morphology, spontaneous intrinsic brain activity, white matter integrity, and global structural network topology, which were in turn related to cognition and emotion in CSVD patients. CONCLUSION: Our findings provide an integrative framework to understand the pathophysiological mechanisms underlying the interplay between gut microbiota dysbiosis and CSVD, highlighting the potential of targeting the microbiota-gut-brain axis as a therapeutic strategy in CSVD patients.

AAAI Conference 2024 Conference Paper

Dynamic Reactive Spiking Graph Neural Network

Han Zhao
Xu Yang
Cheng Deng
Junchi Yan

Spiking Graph Neural Networks are emerging tools for analyzing graph data along with low energy consumption and certain biological fidelity. Existing methods directly integrate same-reactive spiking neurons into graph neural networks for processing propagated graphs. However, such same-reactive neurons are not biological-functionality enough compared to the brain's dynamic-reactive ones, limiting the model's expression. Meanwhile, insufficient long-range neighbor information can be excavated with the few-step propagated graph, restricting discrimination of graph spiking embeddings. Inspired by the dynamic cognition in the brain, we propose a Dynamic Reactive Spiking Graph Neural Network that can enhance model's expressive ability in higher biological fidelity. Specifically, we design dynamic reactive spiking neurons to process spiking graph inputs, which have unique optimizable thresholds to spontaneously explore dynamic reactive states between neurons. Moreover, discriminative graph positional spikes are learned and integrated adaptively into spiking outputs through our neurons, thereby exploring long-range neighbors more thoroughly. Finally, with the dynamic reactive mechanism and learnable positional integration, we can obtain a powerful and highly bio-fidelity model with low energy consumption. Experiments on various domain-related datasets can demonstrate the effectiveness of our model. Our code is available at https://github.com/hzhao98/DRSGNN.

PDF Details DOI

JMLR Journal 2024 Journal Article

Efficient Modality Selection in Multimodal Learning

Yifei He
Runxiang Cheng
Gargi Balasubramaniam
Yao-Hung Hubert Tsai
Han Zhao

Multimodal learning aims to learn from data of different modalities by fusing information from heterogeneous sources. Although it is beneficial to learn from more modalities, it is often infeasible to use all available modalities under limited computational resources. Modeling with all available modalities can also be inefficient and unnecessary when information across input modalities overlaps. In this paper, we study the modality selection problem, which aims to select the most useful subset of modalities for learning under a cardinality constraint. To that end, we propose a unified theoretical framework to quantify the learning utility of modalities, and we identify dependence assumptions to flexibly model the heterogeneous nature of multimodal data, which also allows efficient algorithm design. Accordingly, we derive a greedy modality selection algorithm via submodular maximization, which selects the most useful modalities with an optimality guarantee on learning performance. We also connect marginal-contribution-based feature importance scores, such as Shapley value, from the feature selection domain to the context of modality selection, to efficiently compute the importance of individual modality. We demonstrate the efficacy of our theoretical results and modality selection algorithms on 2 synthetic and 4 real-world data sets on a diverse range of multimodal data. [abs] [ pdf ][ bib ] &copy JMLR 2024. ( edit, beta )

TMLR Journal 2024 Journal Article

Enhancing Compositional Generalization via Compositional Feature Alignment

Haoxiang Wang
Haozhe Si
Huajie Shao
Han Zhao

Real-world applications of machine learning models often confront data distribution shifts, wherein discrepancies exist between the training and test data distributions. In the common multi-domain multi-class setup, as the number of classes and domains scales up, it becomes infeasible to gather training data for every domain-class combination. This challenge naturally leads the quest for models with Compositional Generalization (CG) ability, where models can generalize to unseen domain-class combinations. To delve into the CG challenge, we develop CG-Bench, a suite of CG benchmarks derived from existing real-world image datasets, and observe that the prevalent pretraining-finetuning paradigm on foundational models, such as CLIP and DINOv2, struggles with the challenge. To address this challenge, we propose Compositional Feature Alignment (CFA), a simple two-stage finetuning technique that i) learns two orthogonal linear heads on a pretrained encoder with respect to class and domain labels, and ii) fine-tunes the encoder with the newly learned head frozen. We theoretically and empirically justify that CFA encourages compositional feature learning of pretrained models. We further conduct extensive experiments on CG-Bench for CLIP and DINOv2, two powerful pretrained vision foundation models. Experiment results show that CFA outperforms common finetuning techniques in compositional generalization, corroborating CFA's efficacy in compositional feature learning.

AAAI Conference 2024 Conference Paper

Fair and Optimal Prediction via Post-Processing

Han Zhao

In this talk I will discuss our recent work on characterizing the inherent tradeoff between fairness and accuracy in both classification and regression problems. I will also present a post-processing algorithm that derives optimal fair predictors from Bayes score functions.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

FedGTST: Boosting Global Transferability of Federated Models via Statistics Tuning

Evelyn Ma
Chao Pan
Rasoul Etesami
Han Zhao
Olgica Milenkovic

The performance of Transfer Learning (TL) significantly depends on effective pretraining, which not only requires extensive amounts of data but also substantial computational resources. As a result, in practice, it is challenging to successfully perform TL at the level of individual model developers. Federated Learning (FL) addresses these challenges by enabling collaboration among individual clients through an indirect expansion of the available dataset, distribution of the computation burden across different entities, and privacy-preserving communication mechanisms. Despite several attempts to devise effective transferable FL approaches, several important issues remain unsolved. First, existing methods in this setting primarily focus on optimizing transferability within their local client domains, thereby ignoring transferability over the global learning domain. Second, most approaches focus on analyzing indirect transferability metrics, which does not allow for accurate assessment of the final target loss and extent of transferability. To address these issues, we introduce two important FL features into the model. The first boosts transferability via an exchange protocol between the clients and the server that includes information about cross-client Jacobian (gradient) norms. The second feature promotes an increase of the average of the Jacobians of the clients at the server side, which is subsequently used as a local regularizer that reduces the cross-client Jacobian variance. A rigorous analysis of our transferable federated algorithm, termed FedGTST (Federated Global Transferability via Statistics Tuning), reveals that increasing the averaged Jacobian norm across clients and reducing its variance ensures tight control of the target loss. This insight leads to the first known upper bound on the target loss of transferable federated learning in terms of the source loss and source-target domain discrepancy. Extensive experimental results on datasets including MNIST → MNIST-M and CIFAR10 → SVHN suggest that FedGTST significantly outperforms other relevant baselines, such as FedSR. For example, on the second source-target dataset pair, we improve the accuracy of FedSR by 9. 8% and that of FedIIR by 7. 6% when the backbone used is LeNet.

PDF Details DOI

JMLR Journal 2024 Journal Article

Gradual Domain Adaptation: Theory and Algorithms

Yifei He
Haoxiang Wang
Bo Li
Han Zhao

Unsupervised domain adaptation (UDA) adapts a model from a labeled source domain to an unlabeled target domain in a one-off way. Though widely applied, UDA faces a great challenge whenever the distribution shift between the source and the target is large. Gradual domain adaptation (GDA) mitigates this limitation by using intermediate domains to gradually adapt from the source to the target domain. In this work, we first theoretically analyze gradual self-training, a popular GDA algorithm, and provide a significantly improved generalization bound compared with Kumar et al. (2020). Our theoretical analysis leads to an interesting insight: to minimize the generalization error on the target domain, the sequence of intermediate domains should be placed uniformly along the Wasserstein geodesic between the source and target domains. The insight is particularly useful under the situation where intermediate domains are missing or scarce, which is often the case in real-world applications. Based on the insight, we propose Generative Gradual Domain Adaptation with Optimal Transport (GOAT), an algorithmic framework that can generate intermediate domains in a data-dependent way. More concretely, we first generate intermediate domains along the Wasserstein geodesic between two given consecutive domains in a feature space, then apply gradual self-training to adapt the source-trained classifier to the target along the sequence of intermediate domains. Empirically, we demonstrate that our GOAT framework can improve the performance of standard GDA when the given intermediate domains are scarce, significantly broadening the real-world application scenarios of GDA. Our code is available at https://github.com/uiuctml/GOAT. [abs] [ pdf ][ bib ] [ code ] &copy JMLR 2024. ( edit, beta )

NeurIPS Conference 2024 Conference Paper

Learning Structured Representations with Hyperbolic Embeddings

Aditya Sinha
Siqi Zeng
Makoto Yamada
Han Zhao

Most real-world datasets consist of a natural hierarchy between classes or an inherent label structure that is either already available or can be constructed cheaply. However, most existing representation learning methods ignore this hierarchy, treating labels as permutation invariant. Recent work [Zeng et al. , 2022] proposes using this structured information explicitly, but the use of Euclidean distance may distort the underlying semantic context [Chen et al. , 2013]. In this work, motivated by the advantage of hyperbolic spaces in modeling hierarchical relationships, we propose a novel approach HypStructure: a Hyperbolic Structured regularization approach to accurately embed the label hierarchy into the learned representations. HypStructure is a simple-yet-effective regularizer that consists of a hyperbolic tree-based representation loss along with a centering loss, and can be combined with any standard task loss to learn hierarchy-informed features. Extensive experiments on several large-scale vision benchmarks demonstrate the efficacy of HypStructure in reducing distortion and boosting generalization performance especially under low dimensional scenarios. For a better understanding of structured representation, we perform eigenvalue analysis that links the representation geometry to improved Out-of-Distribution (OOD) detection performance seen empirically.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorch

Xiaoyuan Zhang
Liang Zhao
Yingying Yu
Xi Lin
Yifan Chen
Han Zhao
Qingfu Zhang

Multiobjective optimization problems (MOPs) are prevalent in machine learning, with applications in multi-task learning, learning under fairness or robustness constraints, etc. Instead of reducing multiple objective functions into a scalar objective, MOPs aim to optimize for the so-called Pareto optimality or Pareto set learning, which involves optimizing more than one objective function simultaneously, over models with thousands to millions of parameters. Existing benchmark libraries for MOPs mainly focus on evolutionary algorithms, most of which are zeroth-order or meta-heuristic methods that do not effectively utilize higher-order information from objectives and cannot scale to large-scale models with millions of parameters. In light of the above challenges, this paper introduces \algoname, the first multiobjective optimization library that supports state-of-the-art gradient-based methods, provides a fair and comprehensive benchmark, and is open-sourced for the community.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Most Influential Subset Selection: Challenges, Promises, and Beyond

Yuzheng Hu
Pingbang Hu
Han Zhao
Jiaqi W.

How can we attribute the behaviors of machine learning models to their training data? While the classic influence function sheds light on the impact of individual samples, it often fails to capture the more complex and pronounced collective influence of a set of samples. To tackle this challenge, we study the Most Influential Subset Selection (MISS) problem, which aims to identify a subset of training samples with the greatest collective influence. We conduct a comprehensive analysis of the prevailing approaches in MISS, elucidating their strengths and weaknesses. Our findings reveal that influence-based greedy heuristics, a dominant class of algorithms in MISS, can provably fail even in linear regression. We delineate the failure modes, including the errors of influence function and the non-additive structure of the collective influence. Conversely, we demonstrate that an adaptive version of these heuristics which applies them iteratively, can effectively capture the interactions among samples and thus partially address the issues. Experiments on real-world datasets corroborate these theoretical findings, and further demonstrate that the merit of adaptivity can extend to more complex scenarios such as classification tasks and non-linear neural networks. We conclude our analysis by emphasizing the inherent trade-off between performance and computational efficiency, questioning the use of additive metrics such as the linear datamodeling score, and offering a range of discussions.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

On the Expressive Power of Tree-Structured Probabilistic Circuits

Lang Yin
Han Zhao

Probabilistic circuits (PCs) have emerged as a powerful framework compactly representing probability distributions for efficient and exact probabilistic inference. It has been shown that PCs with general directed acyclic graph (DAG) structure can be understood as a mixture of exponentially (in its height) many components, each of which is a product distributions over univariate marginals. However, existing structure learning algorithms for PCs often generate tree-structured circuits, or using tree-structured circuits as intermediate steps to compress them into DAG-structured circuits. This leads to an intriguing question on whether there exists an exponential gap between DAGs and trees for the PC structure. In this paper, we provide a negative answer to this conjecture by proving that, for $n$ variables, there is a quasi-polynomial upper bound $n^{O(\log n)}$ on the size of an equivalent tree computing the same probability distribution. On the other hand, we will also show that given a depth restriction on the tree, there is a super-polynomial separation between tree and DAG-structured PCs. Our work takes an important step towards understanding the expressive power of tree-structured PCs, and our techniques may be of independent interest in the study of structure learning algorithms for PCs.

PDF Details DOI

TMLR Journal 2024 Journal Article

Personalized Federated Learning with Spurious Features: An Adversarial Approach

Xiaoyang Wang
Han Zhao
Klara Nahrstedt
Sanmi Koyejo

One of the common approaches for personalizing federated learning is fine-tuning the global model for each local client. While this addresses some issues of statistical heterogeneity, we find that such personalization methods are vulnerable to spurious features at local agents, leading to reduced generalization performance. This work considers a setup where spurious features correlate with the label in each client's training environment, and the mixture of multiple training environments (i.e., the global environment) diminishes the spurious correlations. In other words, while the global federated learning model trained over the global environment suffers less from spurious features, the local fine-tuning step may lead to personalized models vulnerable to spurious correlations. In light of this practical and pressing challenge, we propose a novel strategy to mitigate the effect of spurious features during personalization by maintaining the adversarial transferability between the global and personalized models. Empirical results on object and action recognition tasks show that our proposed approach bounds personalized models from further exploiting spurious features while preserving the benefit of enhanced accuracy from fine-tuning.

TMLR Journal 2024 Journal Article

RLHF Workflow: From Reward Modeling to Online RLHF

Hanze Dong
Wei Xiong
Bo Pang
Haoxiang Wang
Han Zhao
Yingbo Zhou
Nan Jiang
Doyen Sahoo

We present the workflow of Online Iterative Reinforcement Learning from Human Feedback (RLHF) in this technical report, which is widely reported to outperform its offline counterpart by a large margin in the recent large language model (LLM) literature. However, existing open-source RLHF projects are still largely confined to the offline learning setting. In this technical report, we aim to fill in this gap and provide a detailed recipe that is easy to reproduce for online iterative RLHF. In particular, since online human feedback is usually infeasible for open-source communities with limited resources, we start by constructing preference models using a diverse set of open-source datasets and use the constructed proxy preference model to approximate human feedback. Then, we discuss the theoretical insights and algorithmic principles behind online iterative RLHF, followed by a detailed practical implementation. Our trained LLM achieves impressive performance on LLM chatbot benchmarks, including AlpacaEval-2, Arena-Hard, and MT-Bench, as well as other academic benchmarks such as HumanEval and TruthfulQA. We have shown that supervised fine-tuning (SFT) and iterative RLHF can obtain state-of-the-art performance with fully open-source datasets. Further, we have made our models, curated datasets, and comprehensive step-by-step code guidebooks publicly available.

JMLR Journal 2023 Journal Article

Adaptation Augmented Model-based Policy Optimization

Jian Shen
Hang Lai
Minghuan Liu
Han Zhao
Yong Yu
Weinan Zhang

Compared to model-free reinforcement learning (RL), model-based RL is often more sample efficient by leveraging a learned dynamics model to help decision making. However, the learned model is usually not perfectly accurate and the error will compound in multi-step predictions, which can lead to poor asymptotic performance. In this paper, we first derive an upper bound of the return discrepancy between the real dynamics and the learned model, which reveals the fundamental problem of distribution shift between simulated data and real data. Inspired by the theoretical analysis, we propose an adaptation augmented model-based policy optimization (AMPO) framework to address the distribution shift problem from the perspectives of feature learning and instance re-weighting, respectively. Specifically, the feature-based variant, namely FAMPO, introduces unsupervised model adaptation to minimize the integral probability metric (IPM) between feature distributions from real and simulated data, while the instance-based variant, termed as IAMPO, utilizes importance sampling to re-weight the real samples used to train the model. Besides model learning, we also investigate how to improve policy optimization in the model usage phase by selecting simulated samples with different probability according to their uncertainty. Extensive experiments on challenging continuous control tasks show that FAMPO and IAMPO, coupled with our model usage technique, achieves superior performance against baselines, which demonstrates the effectiveness of the proposed methods. [abs] [ pdf ][ bib ] &copy JMLR 2023. ( edit, beta )

TMLR Journal 2023 Journal Article

Costs and Benefits of Fair Regression

Han Zhao

Real-world applications of machine learning tools in high-stakes domains are often regulated to be fair, in the sense that the predicted target should satisfy some quantitative notion of parity with respect to a protected attribute. However, the exact tradeoff between fairness and accuracy with a real-valued target is not entirely clear. In this paper, we characterize the inherent tradeoff between statistical parity and accuracy in the regression setting by providing a lower bound on the error of any attribute-blind fair regressor. Our lower bound is sharp, algorithm-independent, and admits a simple interpretation: when the moments of the target differ between groups, any fair algorithm has to make an error on at least one of the groups. We further extend this result to give a lower bound on the joint error of any (approximately) fair algorithm, using the Wasserstein distance to measure the quality of the approximation. With our novel lower bound, we also show that the price paid by a fair regressor that does not take the protected attribute as input is less than that of a fair regressor with explicit access to the protected attribute. On the upside, we establish the first connection between individual fairness, accuracy parity, and the Wasserstein distance by showing that if a regressor is individually fair, it also approximately verifies the accuracy parity, where the gap is again given by the Wasserstein distance between the two groups. Inspired by our theoretical results, we develop a practical algorithm for fair regression through the lens of representation learning, and conduct experiments on a real-world dataset to corroborate our findings.

NeurIPS Conference 2023 Conference Paper

Efficient Learning of Linear Graph Neural Networks via Node Subsampling

Seiyun Shin
Ilan Shomorony
Han Zhao

Graph Neural Networks (GNNs) are a powerful class of machine learning models with applications in recommender systems, drug discovery, social network analysis, and computer vision. One challenge with their implementation is that GNNs often take large-scale graphs as inputs, which imposes significant computational/storage costs in the training and testing phases. In particular, the message passing operations of a GNN require multiplication of the graph adjacency matrix $A \in \mathbb{R}^{n \times n}$ and the data matrix $X \in \mathbb{R}^{n \times d}$, and the $O(n^2 d)$ time complexity can be prohibitive for large $n$. Thus, a natural question is whether it is possible to perform the GNN operations in (quasi-)linear time by avoiding the full computation of $A X$. To study this question, we consider the setting of a regression task on a two-layer Linear Graph Convolutional Network (GCN). We develop an efficient training algorithm based on (1) performing node subsampling, (2) estimating the leverage scores of $A X$ based on the subsampled graph, and (3) performing leverage score sampling on $A X$. We show that our proposed scheme learns the regression model observing only $O(nd\epsilon^{-2}\log n)$ entries of $A$ in time $O(nd^2 \epsilon^{-2}\log n)$, with the guarantee that the learned weights deviate by at most $\epsilon$ under the $\ell_2$ norm from the model learned using the entire adjacency matrix $A$. We present empirical results for regression problems on real-world graphs and show that our algorithm significantly outperforms other baseline sampling strategies that exploit the same number of observations.

AAMAS Conference 2023 Conference Paper

FedMM: A Communication Efficient Solver for Federated Adversarial Domain Adaptation

Yan Shen
Jian Du
Han Zhao
Zhanghexuan Ji
Chunwei Ma
Mingchen Gao

Federated adversary domain adaptation is a unique distributed minimax training task due to the heterogeneous data among different local clients, where each client only sees a subset of the data that merely belongs to either the source or target domain. Despite the extensive research in distributed minimax optimization, existing communication efficient solvers that exploit multiple steps of the local update are still not able to generate satisfactory solutions for federated adversarial domain adaptation because of the gradient divergence issue among clients. To tackle this problem, we propose a distributed minimax optimizer, referred to as FedMM, by introducing dual variables to bridge the gradient gap among clients. This algorithm is effective even in the extreme case where each client has different label classes and some clients only have unlabeled data. We prove that FedMM admits benign convergence to a stationary point under domain-shifted unlabeled data. On a variety of benchmark datasets, extensive experiments show that FedMM consistently achieves both better communication savings and significant accuracy improvements over existing federated optimizers based on the stochastic gradient descent ascent (SGDA) algorithm. When training from scratch, for example, it outperforms other SGDA based federated average methods by around 20% in accuracy over the same communication rounds; and it consistently outperforms when training from pre-trained models.

NeurIPS Conference 2023 Conference Paper

Learning List-Level Domain-Invariant Representations for Ranking

Ruicheng Xian
Honglei Zhuang
Zhen Qin
Hamed Zamani
Jing Lu
Ji Ma
Kai Hui
Han Zhao

Domain adaptation aims to transfer the knowledge learned on (data-rich) source domains to (low-resource) target domains, and a popular method is invariant representation learning, which matches and aligns the data distributions on the feature space. Although this method is studied extensively and applied on classification and regression problems, its adoption on ranking problems is sporadic, and the few existing implementations lack theoretical justifications. This paper revisits invariant representation learning for ranking. Upon reviewing prior work, we found that they implement what we call item-level alignment, which aligns the distributions of the items being ranked from all lists in aggregate but ignores their list structure. However, the list structure should be leveraged, because it is intrinsic to ranking problems where the data and the metrics are defined and computed on lists, not the items by themselves. To close this discrepancy, we propose list-level alignment—learning domain-invariant representations at the higher level of lists. The benefits are twofold: it leads to the first domain adaptation generalization bound for ranking, in turn providing theoretical support for the proposed method, and it achieves better empirical transfer performance for unsupervised domain adaptation on ranking tasks, including passage reranking.

NeurIPS Conference 2023 Conference Paper

Revisiting Scalarization in Multi-Task Learning: A Theoretical Perspective

Yuzheng Hu
Ruicheng Xian
Qilong Wu
Qiuling Fan
Lang Yin
Han Zhao

Linear scalarization, i. e. , combining all loss functions by a weighted sum, has been the default choice in the literature of multi-task learning (MTL) since its inception. In recent years, there is a surge of interest in developing Specialized Multi-Task Optimizers (SMTOs) that treat MTL as a multi-objective optimization problem. However, it remains open whether there is a fundamental advantage of SMTOs over scalarization. In fact, heated debates exist in the community comparing these two types of algorithms, mostly from an empirical perspective. To approach the above question, in this paper, we revisit scalarization from a theoretical perspective. We focus on linear MTL models and study whether scalarization is capable of fully exploring the Pareto front. Our findings reveal that, in contrast to recent works that claimed empirical advantages of scalarization, scalarization is inherently incapable of full exploration, especially for those Pareto optimal solutions that strike the balanced trade-offs between multiple tasks. More concretely, when the model is under-parametrized, we reveal a multi-surface structure of the feasible region and identify necessary and sufficient conditions for full exploration. This leads to the conclusion that scalarization is in general incapable of tracing out the Pareto front. Our theoretical results partially answer the open questions in Xin et al. (2021), and provide a more intuitive explanation on why scalarization fails beyond non-convexity. We additionally perform experiments on a real-world dataset using both scalarization and state-of-the-art SMTOs. The experimental results not only corroborate our theoretical findings, but also unveil the potential of SMTOs in finding balanced solutions, which cannot be achieved by scalarization.

TMLR Journal 2022 Journal Article

Algorithms and Theory for Supervised Gradual Domain Adaptation

Jing Dong
Shiji Zhou
Baoxiang Wang
Han Zhao

The phenomenon of data distribution evolving over time has been observed in a range of applications, calling the needs of adaptive learning algorithms. We thus study the problem of supervised gradual domain adaptation, where labeled data from shifting distributions are available to the learner along the trajectory, and we aim to learn a classifier on a target data distribution of interest. Under this setting, we provide the first generalization upper bound on the learning error under mild assumptions. Our results are algorithm agnostic, general for a range of loss functions, and only depend linearly on the averaged learning error across the trajectory. This shows significant improvement compared to the previous upper bound for unsupervised gradual domain adaptation, where the learning error on the target domain depends exponentially on the initial error on the source domain. Compared with the offline setting of learning from multiple domains, our results also suggest the potential benefits of the temporal structure among different domains in adapting to the target one. Empirically, our theoretical results imply that learning proper representations across the domains will effectively mitigate the learning errors. Motivated by these theoretical insights, we propose a min-max learning objective to learn the representation and classifier simultaneously. Experimental results on both semi-synthetic and large-scale real datasets corroborate our findings and demonstrate the effectiveness of our objectives.

JMLR Journal 2022 Journal Article

Fundamental Limits and Tradeoffs in Invariant Representation Learning

Han Zhao
Chen Dan
Bryon Aragam
Tommi S. Jaakkola
Geoffrey J. Gordon
Pradeep Ravikumar

A wide range of machine learning applications such as privacy-preserving learning, algorithmic fairness, and domain adaptation/generalization among others, involve learning invariant representations of the data that aim to achieve two competing goals: (a) maximize information or accuracy with respect to a target response, and (b) maximize invariance or independence with respect to a set of protected features (e.g.\ for fairness, privacy, etc). Despite their wide applicability, theoretical understanding of the optimal tradeoffs --- with respect to accuracy, and invariance --- achievable by invariant representations is still severely lacking. In this paper, we provide an information theoretic analysis of such tradeoffs under both classification and regression settings. More precisely, we provide a geometric characterization of the accuracy and invariance achievable by any representation of the data; we term this feasible region the information plane. We provide an inner bound for this feasible region for the classification case, and an exact characterization for the regression case, which allows us to either bound or exactly characterize the Pareto optimal frontier between accuracy and invariance. Although our contributions are mainly theoretical, a key practical application of our results is in certifying the potential sub-optimality of any given representation learning algorithm for either classification or regression tasks. Our results shed new light on the fundamental interplay between accuracy and invariance, and may be useful in guiding the design of future representation learning algorithms. [abs] [ pdf ][ bib ] &copy JMLR 2022. ( edit, beta )

JMLR Journal 2022 Journal Article

Inherent Tradeoffs in Learning Fair Representations

Han Zhao
Geoffrey J. Gordon

Real-world applications of machine learning tools in high-stakes domains are often regulated to be fair, in the sense that the predicted target should satisfy some quantitative notion of parity with respect to a protected attribute. However, the exact tradeoff between fairness and accuracy is not entirely clear, even for the basic paradigm of classification problems. In this paper, we characterize an inherent tradeoff between statistical parity and accuracy in the classification setting by providing a lower bound on the sum of group-wise errors of any fair classifiers. Our impossibility theorem could be interpreted as a certain uncertainty principle in fairness: if the base rates differ among groups, then any fair classifier satisfying statistical parity has to incur a large error on at least one of the groups. We further extend this result to give a lower bound on the joint error of any (approximately) fair classifiers, from the perspective of learning fair representations. To show that our lower bound is tight, assuming oracle access to Bayes (potentially unfair) classifiers, we also construct an algorithm that returns a randomized classifier which is both optimal (in terms of accuracy) and fair. Interestingly, when the protected attribute can take more than two values, an extension of this lower bound does not admit an analytic solution. Nevertheless, in this case, we show that the lower bound can be efficiently computed by solving a linear program, which we term as the TV-Barycenter problem, a barycenter problem under the TV-distance. On the upside, we prove that if the group-wise Bayes optimal classifiers are close, then learning fair representations leads to an alternative notion of fairness, known as the accuracy parity, which states that the error rates are close between groups. Finally, we also conduct experiments on real-world datasets to confirm our theoretical findings. [abs] [ pdf ][ bib ] &copy JMLR 2022. ( edit, beta )

AAAI Conference 2022 Conference Paper

Invariant Information Bottleneck for Domain Generalization

Bo Li
Yifei Shen
Yezhen Wang
Wenzhen Zhu
Colorado Reed
Dongsheng Li
Kurt Keutzer
Han Zhao

Invariant risk minimization (IRM) has recently emerged as a promising alternative for domain generalization. Nevertheless, the loss function is difficult to optimize for nonlinear classifiers and the original optimization objective could fail when pseudo-invariant features and geometric skews exist. Inspired by IRM, in this paper we propose a novel formulation for domain generalization, dubbed invariant information bottleneck (IIB). IIB aims at minimizing invariant risks for nonlinear classifiers and simultaneously mitigating the impact of pseudo-invariant features and geometric skews. Specifically, we first present a novel formulation for invariant causal prediction via mutual information. Then we adopt the variational formulation of the mutual information to develop a tractable loss function for nonlinear classifiers. To overcome the failure modes of IRM, we propose to minimize the mutual information between the inputs and the corresponding representations. IIB significantly outperforms IRM on synthetic datasets, where the pseudo-invariant features and geometric skews occur, showing the effectiveness of proposed formulation in overcoming failure modes of IRM. Furthermore, experiments on DomainBed show that IIB outperforms 13 baselines by 0. 9% on average across 7 real datasets.

IJCAI Conference 2021 Conference Paper

Graph Debiased Contrastive Learning with Joint Representation Clustering

Han Zhao
Xu Yang
Zhenru Wang
Erkun Yang
Cheng Deng

By contrasting positive-negative counterparts, graph contrastive learning has become a prominent technique for unsupervised graph representation learning. However, existing methods fail to consider the class information and will introduce false-negative samples in the random negative sampling, causing poor performance. To this end, we propose a graph debiased contrastive learning framework, which can jointly perform representation learning and clustering. Specifically, representations can be optimized by aligning with clustered class information, and simultaneously, the optimized representations can promote clustering, leading to more powerful representations and clustering results. More importantly, we randomly select negative samples from the clusters which are different from the positive sample's cluster. In this way, as the supervisory signals, the clustering results can be utilized to effectively decrease the false-negative samples. Extensive experiments on five datasets demonstrate that our method achieves new state-of-the-art results on graph clustering and classification tasks.

PDF Details DOI

NeurIPS Conference 2021 Conference Paper

Quantifying and Improving Transferability in Domain Generalization

Guojun Zhang
Han Zhao
Yaoliang Yu
Pascal Poupart

Out-of-distribution generalization is one of the key challenges when transferring a model from the lab to the real world. Existing efforts mostly focus on building invariant features among source and target domains. Based on invariant features, a high-performing classifier on source domains could hopefully behave equally well on a target domain. In other words, we hope the invariant features to be \emph{transferable}. However, in practice, there are no perfectly transferable features, and some algorithms seem to learn ``more transferable'' features than others. How can we understand and quantify such \emph{transferability}? In this paper, we formally define transferability that one can quantify and compute in domain generalization. We point out the difference and connection with common discrepancy measures between domains, such as total variation and Wasserstein distance. We then prove that our transferability can be estimated with enough samples and give a new upper bound for the target error based on our transferability. Empirically, we evaluate the transferability of the feature embeddings learned by existing algorithms for domain generalization. Surprisingly, we find that many algorithms are not quite learning transferable features, although few could still survive. In light of this, we propose a new algorithm for learning transferable features and test it over various benchmark datasets, including RotatedMNIST, PACS, Office-Home and WILDS-FMoW. Experimental results show that the proposed algorithm achieves consistent improvement over many state-of-the-art algorithms, corroborating our theoretical findings.

NeurIPS Conference 2020 Conference Paper

Domain Adaptation with Conditional Distribution Matching and Generalized Label Shift

Remi Tachet des Combes
Han Zhao
Yu-Xiang Wang
Geoffrey J. Gordon

Adversarial learning has demonstrated good performance in the unsupervised domain adaptation setting, by learning domain-invariant representations. However, recent work has shown limitations of this approach when label distributions differ between the source and target domains. In this paper, we propose a new assumption, \textit{generalized label shift} ($\glsa$), to improve robustness against mismatched label distributions. $\glsa$ states that, conditioned on the label, there exists a representation of the input that is invariant between the source and target domains. Under $\glsa$, we provide theoretical guarantees on the transfer performance of any classifier. We also devise necessary and sufficient conditions for $\glsa$ to hold, by using an estimation of the relative class weights between domains and an appropriate reweighting of samples. Our weight estimation method could be straightforwardly and generically applied in existing domain adaptation (DA) algorithms that learn domain-invariant representations, with small computational overhead. In particular, we modify three DA algorithms, JAN, DANN and CDAN, and evaluate their performance on standard and artificial DA tasks. Our algorithms outperform the base versions, with vast improvements for large label distribution mismatches. Our code is available at \url{https: //tinyurl. com/y585xt6j}.

NeurIPS Conference 2020 Conference Paper

Model-based Policy Optimization with Unsupervised Model Adaptation

Jian Shen
Han Zhao
Weinan Zhang
Yong Yu

Model-based reinforcement learning methods learn a dynamics model with real data sampled from the environment and leverage it to generate simulated data to derive an agent. However, due to the potential distribution mismatch between simulated data and real data, this could lead to degraded performance. Despite much effort being devoted to reducing this distribution mismatch, existing methods fail to solve it explicitly. In this paper, we investigate how to bridge the gap between real and simulated data due to inaccurate model estimation for better policy optimization. To begin with, we first derive a lower bound of the expected return, which naturally inspires a bound maximization algorithm by aligning the simulated and real data distributions. To this end, we propose a novel model-based reinforcement learning framework AMPO, which introduces unsupervised model adaptation to minimize the integral probability metric (IPM) between feature distributions from real and simulated data. Instantiating our framework with Wasserstein-1 distance gives a practical model-based approach. Empirically, our approach achieves state-of-the-art performance in terms of sample efficiency on a range of continuous control benchmark tasks.

NeurIPS Conference 2020 Conference Paper

Neural Methods for Point-wise Dependency Estimation

Yao-Hung Hubert Tsai
Han Zhao
Makoto Yamada
Louis-Philippe Morency
Russ R. Salakhutdinov

Since its inception, the neural estimation of mutual information (MI) has demonstrated the empirical success of modeling expected dependency between high-dimensional random variables. However, MI is an aggregate statistic and cannot be used to measure point-wise dependency between different events. In this work, instead of estimating the expected dependency, we focus on estimating point-wise dependency (PD), which quantitatively measures how likely two outcomes co-occur. We show that we can naturally obtain PD when we are optimizing MI neural variational bounds. However, optimizing these bounds is challenging due to its large variance in practice. To address this issue, we develop two methods (free of optimizing MI variational bounds): Probabilistic Classifier and Density-Ratio Fitting. We demonstrate the effectiveness of our approaches in 1) MI estimation, 2) self-supervised representation learning, and 3) cross-modal retrieval task.

NeurIPS Conference 2020 Conference Paper

Trade-offs and Guarantees of Adversarial Representation Learning for Information Obfuscation

Han Zhao
Jianfeng Chi
Yuan Tian
Geoffrey J. Gordon

Crowdsourced data used in machine learning services might carry sensitive information about attributes that users do not want to share. Various methods have been proposed to minimize the potential information leakage of sensitive attributes while maximizing the task accuracy. However, little is known about the theory behind these methods. In light of this gap, we develop a novel theoretical framework for attribute obfuscation. Under our framework, we propose a minimax optimization formulation to protect the given attribute and analyze its inference guarantees against worst-case adversaries. Meanwhile, there is a tension between minimizing information leakage and maximizing task accuracy. To understand this, we prove an information-theoretic lower bound to precisely characterize the fundamental trade-off between accuracy and information leakage. We conduct experiments on two real-world datasets to corroborate the inference guarantees and validate the inherent trade-offs therein. Our results indicate that, among several alternatives, the adversarial learning approach achieves the best trade-off in terms of attribute obfuscation and accuracy maximization.

AAMAS Conference 2019 Conference Paper

Deep Generative and Discriminative Domain Adaptation

Han Zhao
Junjie Hu
Zhenyao Zhu
Adam Coates
Geoff Gordon

The ability to adapt to and learn from different domains and environments is crucial for agents to generalize. In this paper we propose a probabilistic framework for domain adaptation that blends both generative and discriminative modeling in a principled way. Under this framework, generative and discriminative models correspond to specific choices of the prior over parameters. By maximizing both the marginal and the conditional log-likelihoods, our models can use both labeled instances from the source domain as well as unlabeled instances from both source and target domains. We show that the popular reconstruction loss of autoencoder corresponds to an upper bound of the negative marginal log-likelihoods of unlabeled instances, and give a generalization bound that explicitly incorporates it into the analysis. We instantiate our framework using neural networks, and build a concrete model, DAuto.

NeurIPS Conference 2019 Conference Paper

Inherent Tradeoffs in Learning Fair Representations

Han Zhao
Geoff Gordon

With the prevalence of machine learning in high-stakes applications, especially the ones regulated by anti-discrimination laws or societal norms, it is crucial to ensure that the predictive models do not propagate any existing bias or discrimination. Due to the ability of deep neural nets to learn rich representations, recent advances in algorithmic fairness have focused on learning fair representations with adversarial techniques to reduce bias in data while preserving utility simultaneously. In this paper, through the lens of information theory, we provide the first result that quantitatively characterizes the tradeoff between demographic parity and the joint utility across different population groups. Specifically, when the base rates differ between groups, we show that any method aiming to learn fair representations admits an information-theoretic lower bound on the joint error across these groups. To complement our negative results, we also prove that if the optimal decision functions across different groups are close, then learning fair representations leads to an alternative notion of fairness, known as the accuracy parity, which states that the error rates are close between groups. Finally, our theoretical findings are also confirmed empirically on real-world datasets.

NeurIPS Conference 2019 Conference Paper

Learning Neural Networks with Adaptive Regularization

Han Zhao
Yao-Hung Hubert Tsai
Russ Salakhutdinov
Geoffrey Gordon

Feed-forward neural networks can be understood as a combination of an intermediate representation and a linear hypothesis. While most previous works aim to diversify the representations, we explore the complementary direction by performing an adaptive and data-dependent regularization motivated by the empirical Bayes method. Specifically, we propose to construct a matrix-variate normal prior (on weights) whose covariance matrix has a Kronecker product structure. This structure is designed to capture the correlations in neurons through backpropagation. Under the assumption of this Kronecker factorization, the prior encourages neurons to borrow statistical strength from one another. Hence, it leads to an adaptive and data-dependent regularization when training networks on small datasets. To optimize the model, we present an efficient block coordinate descent algorithm with analytical solutions. Empirically, we demonstrate that the proposed method helps networks converge to local optima with smaller stable ranks and spectral norms. These properties suggest better generalizations and we present empirical results to support this expectation. We also verify the effectiveness of the approach on multiclass classification and multitask regression problems with various network structures. Our code is publicly available at: ~\url{https: //github. com/yaohungt/Adaptive-Regularization-Neural-Network}.

IJCAI Conference 2019 Conference Paper

On Strategyproof Conference Peer Review

Yichong Xu
Han Zhao
Xiaofei Shi
Nihar B. Shah

We consider peer review under a conference setting where there are conflicts between the reviewers and the submissions. Under such conflicts, reviewers can manipulate their reviews in a strategic manner to influence the final rankings of their own papers. Present-day peer-review systems are not designed to guard against such strategic behavior, beyond minimal (and insufficient) checks such as not assigning a paper to a conflicted reviewer. In this work, we address this problem through the lens of social choice, and present a theoretical framework for strategyproof and efficient peer review. Given the conflict graph which satisfies a simple property, we first present and analyze a flexible framework for reviewer-assignment and aggregation for the reviews that guarantees not only strategyproofness but also a natural efficiency property (unanimity). Our framework is based on the so-called partitioning method, and can be treated as a generalization of this type of method to conference peer review settings. We then empirically show that the requisite property on the (authorship) conflict graph is indeed satisfied in the ICLR-17 submissions data, and further demonstrate a simple trick to make the partitioning method more practically appealing under conference peer-review settings. Finally, we complement our positive results with negative theoretical results where we prove that under slightly stronger requirements, it is impossible for any algorithm to be both strategyproof and efficient.

NeurIPS Conference 2018 Conference Paper

Adversarial Multiple Source Domain Adaptation

Han Zhao
Shanghang Zhang
Guanhang Wu
José M. F. Moura
Joao Costeira
Geoffrey Gordon

While domain adaptation has been actively researched, most algorithms focus on the single-source-single-target adaptation setting. In this paper we propose new generalization bounds and algorithms under both classification and regression settings for unsupervised multiple source domain adaptation. Our theoretical analysis naturally leads to an efficient learning strategy using adversarial neural networks: we show how to interpret it as learning feature representations that are invariant to the multiple domain shifts while still being discriminative for the learning task. To this end, we propose multisource domain adversarial networks (MDAN) that approach domain adaptation by optimizing task-adaptive generalization bounds. To demonstrate the effectiveness of MDAN, we conduct extensive experiments showing superior adaptation performance on both classification and regression problems: sentiment analysis, digit classification, and vehicle counting.

NeurIPS Conference 2017 Conference Paper

Linear Time Computation of Moments in Sum-Product Networks

Han Zhao
Geoffrey Gordon

Bayesian online algorithms for Sum-Product Networks (SPNs) need to update their posterior distribution after seeing one single additional instance. To do so, they must compute moments of the model parameters under this distribution. The best existing method for computing such moments scales quadratically in the size of the SPN, although it scales linearly for trees. This unfortunate scaling makes Bayesian online algorithms prohibitively expensive, except for small or tree-structured SPNs. We propose an optimal linear-time algorithm that works even when the SPN is a general directed acyclic graph (DAG), which significantly broadens the applicability of Bayesian online algorithms for SPNs. There are three key ingredients in the design and analysis of our algorithm: 1). For each edge in the graph, we construct a linear time reduction from the moment computation problem to a joint inference problem in SPNs. 2). Using the property that each SPN computes a multilinear polynomial, we give an efficient procedure for polynomial evaluation by differentiation without expanding the network that may contain exponentially many monomials. 3). We propose a dynamic programming method to further reduce the computation of the moments of all the edges in the graph from quadratic to linear. We demonstrate the usefulness of our linear time algorithm by applying it to develop a linear time assume density filter (ADF) for SPNs.

AAAI Conference 2017 Conference Paper

Unsupervised Domain Adaptation with a Relaxed Covariate Shift Assumption

Tameem Adel
Han Zhao
Alexander Wong

Domain adaptation addresses learning tasks where training is performed on data from one domain whereas testing is performed on data belonging to a different but related domain. Assumptions about the relationship between the source and target domains should lead to tractable solutions on the one hand, and be realistic on the other hand. Here we propose a generative domain adaptation model that allows for modelling different assumptions about this relationship, among which is a newly introduced assumption that replaces covariate shift with a possibly more realistic assumption without losing tractability due to the efﬁcient variational inference procedure developed. In addition to the ability to model less restrictive relationships between source and target, modelling can be performed without any target labeled data (unsupervised domain adaptation). We also provide a Rademacher complexity bound of the proposed algorithm. We evaluate the model on the Amazon reviews and the CVC pedestrian detection datasets.

NeurIPS Conference 2016 Conference Paper

A Unified Approach for Learning the Parameters of Sum-Product Networks

Han Zhao
Pascal Poupart
Geoffrey Gordon

We present a unified approach for learning the parameters of Sum-Product networks (SPNs). We prove that any complete and decomposable SPN is equivalent to a mixture of trees where each tree corresponds to a product of univariate distributions. Based on the mixture model perspective, we characterize the objective function when learning SPNs based on the maximum likelihood estimation (MLE) principle and show that the optimization problem can be formulated as a signomial program. We construct two parameter learning algorithms for SPNs by using sequential monomial approximations (SMA) and the concave-convex procedure (CCCP), respectively. The two proposed methods naturally admit multiplicative updates, hence effectively avoiding the projection operation. With the help of the unified framework, we also show that, in the case of SPNs, CCCP leads to the same algorithm as Expectation Maximization (EM) despite the fact that they are different in general.

IJCAI Conference 2015 Conference Paper

Self-Adaptive Hierarchical Sentence Model

Han Zhao
Zhengdong Lu
Pascal Poupart

The ability to accurately model a sentence at varying stages (e. g. , word-phrase-sentence) plays a central role in natural language processing. As an effort towards this goal we propose a selfadaptive hierarchical sentence model (AdaSent). AdaSent effectively forms a hierarchy of representations from words to phrases and then to sentences through recursive gated local composition of adjacent segments. We design a competitive mechanism (through gating networks) to allow the representations of the same sentence to be engaged in a particular learning task (e. g. , classification), therefore effectively mitigating the gradient vanishing problem persistent in other recursive models. Both qualitative and quantitative analysis shows that AdaSent can automatically form and select the representations suitable for the task at hand during training, yielding superior classification performance over competitor models on 5 benchmark data sets.

AAAI Conference 2015 Conference Paper

SoF: Soft-Cluster Matrix Factorization for Probabilistic Clustering

Han Zhao
Pascal Poupart
Yongfeng Zhang
Martin Lysy

We propose SoF (Soft-cluster matrix Factorization), a probabilistic clustering algorithm which softly assigns each data point into clusters. Unlike model-based clustering algorithms, SoF does not make assumptions about the data density distribution. Instead, we take an axiomatic approach to define 4 properties that the probability of co-clustered pairs of points should satisfy. Based on the properties, SoF utilizes a distance measure between pairs of points to induce the conditional co-cluster probabilities. The objective function in our framework establishes an important connection between probabilistic clustering and constrained symmetric Nonnegative Matrix Factorization (NMF), hence providing a theoretical interpretation for NMF-based clustering algorithms. To optimize the objective, we derive a sequential minimization algorithm using a penalty method. Experimental results on both synthetic and real-world datasets show that SoF significantly outperforms previous NMF-based algorithms and that it is able to detect non-convex patterns as well as cluster boundaries.