Search Results for author: Jun Zhu

Found 267 papers, 111 papers with code

Understanding and Stabilizing GANs' Training Dynamics Using Control Theory

no code implementations ICML 2020 Kun Xu, Chongxuan Li, Jun Zhu, Bo Zhang

There are existing efforts that model the training dynamics of GANs in the parameter space but the analysis cannot directly motivate practically effective stabilizing methods.

Defense Against Adversarial Attacks via Controlling Gradient Leaking on Embedded Manifolds

no code implementations ECCV 2020 Yueru Li, Shuyu Cheng, Hang Su, Jun Zhu

Based on our investigation, we further present a new robust learning algorithm which encourages a larger gradient component in the tangent space of data manifold, suppressing the gradient leaking phenomenon consequently.

Variance Reduction and Quasi-Newton for Particle-Based Variational Inference

no code implementations ICML 2020 Michael Zhu, Chang Liu, Jun Zhu

Particle-based Variational Inference methods (ParVIs), like Stein Variational Gradient Descent, are nonparametric variational inference methods that optimize a set of particles to best approximate a target distribution.

Bayesian Inference Riemannian optimization +1

A Comprehensive Survey of Continual Learning: Theory, Method and Application

no code implementations31 Jan 2023 Liyuan Wang, Xingxing Zhang, Hang Su, Jun Zhu

To cope with real-world dynamics, an intelligent agent needs to incrementally acquire, update, accumulate, and exploit knowledge throughout its lifetime.

Continual Learning Learning Theory

Semiparametric Regression for Spatial Data via Deep Learning

no code implementations10 Jan 2023 Kexuan Li, Jun Zhu, Anthony R. Ives, Volker C. Radeloff, Fangfang Wang

To be specific, we use a sparsely connected deep neural network with rectified linear unit (ReLU) activation function to estimate the unknown regression function that describes the relationship between response and covariates in the presence of spatial dependence.

regression

Why Are Conditional Generative Models Better Than Unconditional Ones?

no code implementations1 Dec 2022 Fan Bao, Chongxuan Li, Jiacheng Sun, Jun Zhu

Extensive empirical evidence demonstrates that conditional generative models are easier to train and perform better than unconditional ones by exploiting the labels of data.

DQ-DETR: Dual Query Detection Transformer for Phrase Extraction and Grounding

1 code implementation28 Nov 2022 Shilong Liu, Yaoyuan Liang, Feng Li, Shijia Huang, Hao Zhang, Hang Su, Jun Zhu, Lei Zhang

As phrase extraction can be regarded as a $1$D text segmentation problem, we formulate PEG as a dual detection problem and propose a novel DQ-DETR model, which introduces dual queries to probe different features from image and text for object prediction and phrase mask prediction.

object-detection Object Detection +4

Physics-Informed Machine Learning: A Survey on Problems, Methods and Applications

no code implementations15 Nov 2022 Zhongkai Hao, Songming Liu, Yichi Zhang, Chengyang Ying, Yao Feng, Hang Su, Jun Zhu

Recent work shows that it provides potential benefits for machine learning models by incorporating the physical prior and collected data, which makes the intersection of machine learning and physics become a prevailing paradigm.

Physics-informed machine learning

DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models

1 code implementation2 Nov 2022 Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, Jun Zhu

The commonly-used fast sampler for guided sampling is DDIM, a first-order diffusion ODE solver that generally needs 100 to 250 steps for high-quality samples.

Text to image generation Text-to-Image Generation

Improving transferability of 3D adversarial attacks with scale and shear transformations

no code implementations2 Nov 2022 Jinali Zhang, Yinpeng Dong, Jun Zhu, Jihong Zhu, Minchi Kuang, Xiaming Yuan

Extensive experiments show that the SS attack proposed in this paper can be seamlessly combined with the existing state-of-the-art (SOTA) 3D point cloud attack methods to form more powerful attack methods, and the SS attack improves the transferability over 3. 6 times compare to the baseline.

Model-based Reinforcement Learning with a Hamiltonian Canonical ODE Network

no code implementations2 Nov 2022 Yao Feng, Yuhong Jiang, Hang Su, Dong Yan, Jun Zhu

Model-based reinforcement learning usually suffers from a high sample complexity in training the world model, especially for the environments with complex dynamics.

Model-based Reinforcement Learning reinforcement-learning +1

Spectral Representation Learning for Conditional Moment Models

no code implementations29 Oct 2022 Ziyu Wang, Yucen Luo, Yueru Li, Jun Zhu, Bernhard Schölkopf

For nonparametric conditional moment models, efficient estimation often relies on preimposed conditions on various measures of ill-posedness of the hypothesis space, which are hard to validate when flexible models are used.

Causal Inference Representation Learning

Isometric 3D Adversarial Examples in the Physical World

no code implementations27 Oct 2022 Yibo Miao, Yinpeng Dong, Jun Zhu, Xiao-Shan Gao

For naturalness, we constrain the adversarial example to be $\epsilon$-isometric to the original one by adopting the Gaussian curvature as a surrogate metric guaranteed by a theoretical analysis.

Neural Eigenfunctions Are Structured Representation Learners

1 code implementation23 Oct 2022 Zhijie Deng, Jiaxin Shi, Hao Zhang, Peng Cui, Cewu Lu, Jun Zhu

In this paper, we introduce a scalable method for learning structured, adaptive-length deep representations.

Contrastive Learning Feature Importance +6

Accelerated Linearized Laplace Approximation for Bayesian Deep Learning

1 code implementation23 Oct 2022 Zhijie Deng, Feng Zhou, Jun Zhu

Laplace approximation (LA) and its linearized variant (LLA) enable effortless adaptation of pretrained deep neural networks to Bayesian neural networks.

A Unified Hard-Constraint Framework for Solving Geometrically Complex PDEs

1 code implementation6 Oct 2022 Songming Liu, Zhongkai Hao, Chengyang Ying, Hang Su, Jun Zhu, Ze Cheng

We present a unified hard-constraint framework for solving geometrically complex PDEs with neural networks, where the most commonly used Dirichlet, Neumann, and Robin boundary conditions (BCs) are considered.

Equivariant Energy-Guided SDE for Inverse Molecular Design

2 code implementations30 Sep 2022 Fan Bao, Min Zhao, Zhongkai Hao, Peiyao Li, Chongxuan Li, Jun Zhu

Inverse molecular design is critical in material science and drug discovery, where the generated molecules should satisfy certain desirable properties.

Drug Discovery

INT: Towards Infinite-frames 3D Detection with An Efficient Framework

no code implementations30 Sep 2022 Jianyun Xu, Zhenwei Miao, Da Zhang, Hongyu Pan, Kaixuan Liu, Peihan Hao, Jun Zhu, Zhengyang Sun, Hongmin Li, Xin Zhan

By employing INT on CenterPoint, we can get around 7% (Waymo) and 15% (nuScenes) performance boost with only 2~4ms latency overhead, and currently SOTA on the Waymo 3D Detection leaderboard.

Offline Reinforcement Learning via High-Fidelity Generative Behavior Modeling

no code implementations29 Sep 2022 Huayu Chen, Cheng Lu, Chengyang Ying, Hang Su, Jun Zhu

To address this problem, we adopt a generative approach by decoupling the learned policy into two parts: an expressive generative behavior model and an action evaluation model.

D4RL Offline RL +2

All are Worth Words: A ViT Backbone for Diffusion Models

2 code implementations25 Sep 2022 Fan Bao, Shen Nie, Kaiwen Xue, Yue Cao, Chongxuan Li, Hang Su, Jun Zhu

In particular, a latent diffusion model with a small U-ViT achieves a record-breaking FID of 5. 48 in text-to-image generation on MS-COCO, among methods without accessing large external datasets during the training of generative models.

Conditional Image Generation Text to image generation +1

Bi-level Physics-Informed Neural Networks for PDE Constrained Optimization using Broyden's Hypergradients

no code implementations15 Sep 2022 Zhongkai Hao, Chengyang Ying, Hang Su, Jun Zhu, Jian Song, Ze Cheng

In this paper, we present a novel bi-level optimization framework to resolve the challenge by decoupling the optimization of the targets and constraints.

On the Reuse Bias in Off-Policy Reinforcement Learning

no code implementations15 Sep 2022 Chengyang Ying, Zhongkai Hao, Xinning Zhou, Hang Su, Dong Yan, Jun Zhu

In this paper, we reveal that the instability is also related to a new notion of Reuse Bias of IS -- the bias in off-policy evaluation caused by the reuse of the replay buffer for evaluation and optimization.

Continuous Control Off-policy evaluation +1

Regret Analysis for Hierarchical Experts Bandit Problem

no code implementations11 Aug 2022 Qihan Guo, Siwei Wang, Jun Zhu

We study an extension of standard bandit problem in which there are R layers of experts.

Robust Learning of Deep Time Series Anomaly Detection Models with Contaminated Training Data

no code implementations3 Aug 2022 Wenkai Li, Cheng Feng, Ting Chen, Jun Zhu

In this work, to tackle this important challenge, we firstly investigate the robustness of commonly used deep TSAD methods with contaminated training data which provides a guideline for applying these methods when the provided training data are not guaranteed to be anomaly-free.

Anomaly Detection Time Series Anomaly Detection

EGSDE: Unpaired Image-to-Image Translation via Energy-Guided Stochastic Differential Equations

1 code implementation14 Jul 2022 Min Zhao, Fan Bao, Chongxuan Li, Jun Zhu

Further, we provide an alternative explanation of the EGSDE as a product of experts, where each of the three experts (corresponding to the SDE and two feature extractors) solely contributes to faithfulness or realism.

Image-to-Image Translation Translation

Thompson Sampling for (Combinatorial) Pure Exploration

no code implementations18 Jun 2022 Siwei Wang, Jun Zhu

To make the algorithm efficient, they usually use the sum of upper confidence bounds within arm set $S$ to represent the upper confidence bound of $S$, which can be much larger than the tight upper confidence bound of $S$ and leads to a much higher complexity than necessary, since the empirical means of different arms in $S$ are independent.

Thompson Sampling

Fast Lossless Neural Compression with Integer-Only Discrete Flows

1 code implementation17 Jun 2022 Siyu Wang, Jianfei Chen, Chongxuan Li, Jun Zhu, Bo Zhang

In this work, we propose Integer-only Discrete Flows (IODF), an efficient neural compressor with integer-only arithmetic.

Quantization

Maximum Likelihood Training for Score-Based Diffusion ODEs by High-Order Denoising Score Matching

1 code implementation16 Jun 2022 Cheng Lu, Kaiwen Zheng, Fan Bao, Jianfei Chen, Chongxuan Li, Jun Zhu

To fill up this gap, we show that the negative likelihood of the ODE can be bounded by controlling the first, second, and third-order score matching errors; and we further present a novel high-order denoising score matching method to enable maximum likelihood training of score-based diffusion ODEs.

Denoising

Estimating the Optimal Covariance with Imperfect Mean in Diffusion Probabilistic Models

1 code implementation15 Jun 2022 Fan Bao, Chongxuan Li, Jiacheng Sun, Jun Zhu, Bo Zhang

Thus, the generation performance on a subset of timesteps is crucial, which is greatly influenced by the covariance design in DPMs.

Consistent Attack: Universal Adversarial Perturbation on Embodied Vision Navigation

no code implementations12 Jun 2022 You Qiaoben, Chengyang Ying, Xinning Zhou, Hang Su, Jun Zhu, Bo Zhang

However, deep neural networks are vulnerable to malicious adversarial noises, which may potentially cause catastrophic failures in Embodied Vision Navigation.

Diagnosing Ensemble Few-Shot Classifiers

no code implementations9 Jun 2022 Weikai Yang, Xi Ye, Xingxing Zhang, Lanxi Xiao, Jiazhi Xia, Zhongyuan Wang, Jun Zhu, Hanspeter Pfister, Shixia Liu

The base learners and labeled samples (shots) in an ensemble few-shot classifier greatly affect the model performance.

GSmooth: Certified Robustness against Semantic Transformations via Generalized Randomized Smoothing

no code implementations9 Jun 2022 Zhongkai Hao, Chengyang Ying, Yinpeng Dong, Hang Su, Jun Zhu, Jian Song

Under the GSmooth framework, we present a scalable algorithm that uses a surrogate image-to-image network to approximate the complex transformation.

Towards Safe Reinforcement Learning via Constraining Conditional Value-at-Risk

no code implementations9 Jun 2022 Chengyang Ying, Xinning Zhou, Hang Su, Dong Yan, Ning Chen, Jun Zhu

Though deep reinforcement learning (DRL) has obtained substantial success, it may encounter catastrophic failures due to the intrinsic uncertainty of both transition and observation.

Continuous Control reinforcement-learning +2

BadDet: Backdoor Attacks on Object Detection

no code implementations28 May 2022 Shih-Han Chan, Yinpeng Dong, Jun Zhu, Xiaolu Zhang, Jun Zhou

We propose four kinds of backdoor attacks for object detection task: 1) Object Generation Attack: a trigger can falsely generate an object of the target class; 2) Regional Misclassification Attack: a trigger can change the prediction of a surrounding object to the target class; 3) Global Misclassification Attack: a single trigger can change the predictions of all objects in an image to the target class; and 4) Object Disappearance Attack: a trigger can make the detector fail to detect the object of the target class.

Autonomous Driving Backdoor Attack +3

Censored Quantile Regression Neural Networks for Distribution-Free Survival Analysis

1 code implementation26 May 2022 Tim Pearce, Jong-Hyeon Jeong, Yichen Jia, Jun Zhu

To offer theoretical insight into our algorithm, we show firstly that it can be interpreted as a form of expectation-maximisation, and secondly that it exhibits a desirable `self-correcting' property.

regression Survival Analysis

Fast Instrument Learning with Faster Rates

1 code implementation22 May 2022 Ziyu Wang, Yuhao Zhou, Jun Zhu

We investigate nonlinear instrumental variable (IV) regression given high-dimensional instruments.

Model Selection regression

Towards Job-Transition-Tag Graph for a Better Job Title Representation Learning

1 code implementation Findings (NAACL) 2022 Jun Zhu, Céline Hudelot

Works on learning job title representation are mainly based on \textit{Job-Transition Graph}, built from the working history of talents.

Representation Learning TAG

Deep Ensemble as a Gaussian Process Approximate Posterior

no code implementations30 Apr 2022 Zhijie Deng, Feng Zhou, Jianfei Chen, Guoqiang Wu, Jun Zhu

In this way, we relate DE to Bayesian inference to enjoy reliable Bayesian uncertainty.

Bayesian Inference

NeuralEF: Deconstructing Kernels by Deep Neural Networks

1 code implementation30 Apr 2022 Zhijie Deng, Jiaxin Shi, Jun Zhu

Learning the principal eigenfunctions of an integral operator defined by a kernel and a data distribution is at the core of many machine learning problems.

Image Classification

Policy Learning for Robust Markov Decision Process with a Mismatched Generative Model

no code implementations13 Mar 2022 Jialian Li, Tongzheng Ren, Dong Yan, Hang Su, Jun Zhu

Our goal is to identify a near-optimal robust policy for the perturbed testing environment, which introduces additional technical difficulties as we need to simultaneously estimate the training environment uncertainty from samples and find the worst-case perturbation for testing.

Query-Efficient Black-box Adversarial Attacks Guided by a Transfer-based Prior

1 code implementation13 Mar 2022 Yinpeng Dong, Shuyu Cheng, Tianyu Pang, Hang Su, Jun Zhu

However, the existing methods inevitably suffer from low attack success rates or poor query efficiency since it is difficult to estimate the gradient in a high-dimensional input space with limited information.

Controllable Evaluation and Generation of Physical Adversarial Patch on Face Recognition

no code implementations9 Mar 2022 Xiao Yang, Yinpeng Dong, Tianyu Pang, Zihao Xiao, Hang Su, Jun Zhu

It is therefore imperative to develop a framework that can enable a comprehensive evaluation of the vulnerability of face recognition in the physical world.

3D Face Modelling Face Recognition

DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection

8 code implementations7 Mar 2022 Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M. Ni, Heung-Yeung Shum

Compared to other models on the leaderboard, DINO significantly reduces its model size and pre-training data size while achieving better results.

 Ranked #1 on Object Detection on COCO 2017 val (box AP metric)

Real-Time Object Detection

Robustness and Accuracy Could Be Reconcilable by (Proper) Definition

1 code implementation21 Feb 2022 Tianyu Pang, Min Lin, Xiao Yang, Jun Zhu, Shuicheng Yan

The trade-off between robustness and accuracy has been widely studied in the adversarial literature.

Inductive Bias

Memory Replay with Data Compression for Continual Learning

1 code implementation ICLR 2022 Liyuan Wang, Xingxing Zhang, Kuo Yang, Longhui Yu, Chongxuan Li, Lanqing Hong, Shifeng Zhang, Zhenguo Li, Yi Zhong, Jun Zhu

In this work, we propose memory replay with data compression (MRDC) to reduce the storage cost of old training samples and thus increase their amount that can be stored in the memory buffer.

Autonomous Driving class-incremental learning +5

DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR

4 code implementations ICLR 2022 Shilong Liu, Feng Li, Hao Zhang, Xiao Yang, Xianbiao Qi, Hang Su, Jun Zhu, Lei Zhang

We present in this paper a novel query formulation using dynamic anchor boxes for DETR (DEtection TRansformer) and offer a deeper understanding of the role of queries in DETR.

Object Detection

Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models

2 code implementations ICLR 2022 Fan Bao, Chongxuan Li, Jun Zhu, Bo Zhang

In this work, we present a surprising result that both the optimal reverse variance and the corresponding optimal KL divergence of a DPM have analytic forms w. r. t.

BE-STI: Spatial-Temporal Integrated Network for Class-Agnostic Motion Prediction With Bidirectional Enhancement

no code implementations CVPR 2022 Yunlong Wang, Hongyu Pan, Jun Zhu, Yu-Huan Wu, Xin Zhan, Kun Jiang, Diange Yang

In this paper, we propose a novel Spatial-Temporal Integrated network with Bidirectional Enhancement, BE-STI, to improve the temporal motion prediction performance by spatial semantic features, which points out an efficient way to combine semantic segmentation and motion prediction.

Autonomous Driving motion prediction +1

AutoLoss-GMS: Searching Generalized Margin-Based Softmax Loss Function for Person Re-Identification

no code implementations CVPR 2022 Hongyang Gu, Jianmin Li, Guangyuan Fu, Chifong Wong, Xinghao Chen, Jun Zhu

In this paper, we propose a novel method, AutoLoss-GMS, to search the better loss function in the space of generalized margin-based softmax loss function for person re-identification automatically.

Person Re-Identification

AFEC: Active Forgetting of Negative Transfer in Continual Learning

1 code implementation NeurIPS 2021 Liyuan Wang, Mingtian Zhang, Zhongfan Jia, Qian Li, Chenglong Bao, Kaisheng Ma, Jun Zhu, Yi Zhong

Without accessing to the old training samples, knowledge transfer from the old tasks to each new task is difficult to determine, which might be either positive or negative.

Continual Learning Transfer Learning

Model-Agnostic Meta-Attack: Towards Reliable Evaluation of Adversarial Robustness

no code implementations13 Oct 2021 Xiao Yang, Yinpeng Dong, Wenzhao Xiang, Tianyu Pang, Hang Su, Jun Zhu

The vulnerability of deep neural networks to adversarial examples has motivated an increasing number of defense strategies for promoting model robustness.

Adversarial Robustness

TiKick: Towards Playing Multi-agent Football Full Games from Single-agent Demonstrations

1 code implementation9 Oct 2021 Shiyu Huang, Wenze Chen, Longfei Zhang, Shizhen Xu, Ziyang Li, Fengming Zhu, Deheng Ye, Ting Chen, Jun Zhu

To the best of our knowledge, Tikick is the first learning-based AI system that can take over the multi-agent Google Research Football full game, while previous work could either control a single agent or experiment on toy academic scenarios.

Starcraft Starcraft II

Ranking Cost: Building An Efficient and Scalable Circuit Routing Planner with Evolution-Based Optimization

1 code implementation8 Oct 2021 Shiyu Huang, Bin Wang, Dong Li, Jianye Hao, Ting Chen, Jun Zhu

In this work, we propose a new algorithm for circuit routing, named Ranking Cost, which innovatively combines search-based methods (i. e., A* algorithm) and learning-based methods (i. e., Evolution Strategies) to form an efficient and trainable router.

Adversarial Semantic Contour for Object Detection

no code implementations ICML Workshop AML 2021 Yichi Zhang, Zijian Zhu, Xiao Yang, Jun Zhu

To address this issue, we propose a novel method of Adversarial Semantic Contour (ASC) guided by object contour as prior.

object-detection Object Detection

Deep Ensemble as a Gaussian Process Posterior

no code implementations29 Sep 2021 Zhijie Deng, Feng Zhou, Jianfei Chen, Guoqiang Wu, Jun Zhu

Deep Ensemble (DE) is a flexible, feasible, and effective alternative to Bayesian neural networks (BNNs) for uncertainty estimation in deep learning.

Variational Inference

Regularized-OFU: an efficient algorithm for general contextual bandit with optimization oracles

no code implementations29 Sep 2021 Yichi Zhou, Shihong Song, Huishuai Zhang, Jun Zhu, Wei Chen, Tie-Yan Liu

In contextual bandit, one major challenge is to develop theoretically solid and empirically efficient algorithms for general function classes.

Multi-Armed Bandits Thompson Sampling

Cluster Attack: Query-based Adversarial Attacks on Graphs with Graph-Dependent Priors

1 code implementation ICML Workshop AML 2021 Zhengyi Wang, Zhongkai Hao, Ziqiao Wang, Hang Su, Jun Zhu

In this work, we propose Cluster Attack -- a Graph Injection Attack (GIA) on node classification, which injects fake nodes into the original graph to degenerate the performance of graph neural networks (GNNs) on certain victim nodes while affecting the other nodes as little as possible.

Adversarial Attack Graph Clustering +2

Tianshou: a Highly Modularized Deep Reinforcement Learning Library

1 code implementation29 Jul 2021 Jiayi Weng, Huayu Chen, Dong Yan, Kaichao You, Alexis Duburcq, Minghao Zhang, Yi Su, Hang Su, Jun Zhu

In this paper, we present Tianshou, a highly modularized Python library for deep reinforcement learning (DRL) that uses PyTorch as its backend.

reinforcement-learning reinforcement Learning

Query2Label: A Simple Transformer Way to Multi-Label Classification

2 code implementations22 Jul 2021 Shilong Liu, Lei Zhang, Xiao Yang, Hang Su, Jun Zhu

The use of Transformer is rooted in the need of extracting local discriminative features adaptively for different labels, which is a strongly desired property due to the existence of multiple objects in one image.

Classification Multi-Label Classification

On the Convergence of Prior-Guided Zeroth-Order Optimization Algorithms

1 code implementation NeurIPS 2021 Shuyu Cheng, Guoqiang Wu, Jun Zhu

Finally, our theoretical results are confirmed by experiments on several numerical benchmarks as well as adversarial attacks.

Understanding Adversarial Attacks on Observations in Deep Reinforcement Learning

no code implementations30 Jun 2021 You Qiaoben, Chengyang Ying, Xinning Zhou, Hang Su, Jun Zhu, Bo Zhang

In this paper, we provide a framework to better understand the existing methods by reformulating the problem of adversarial attacks on reinforcement learning in the function space.

reinforcement-learning reinforcement Learning

Improving Transferability of Adversarial Patches on Face Recognition with Generative Models

no code implementations CVPR 2021 Zihao Xiao, Xianfeng Gao, Chilin Fu, Yinpeng Dong, Wei Gao, Xiaolu Zhang, Jun Zhou, Jun Zhu

However, deep CNNs are vulnerable to adversarial patches, which are physically realizable and stealthy, raising new security concerns on the real-world applications of these models.

Face Recognition

Regularized OFU: an Efficient UCB Estimator forNon-linear Contextual Bandit

no code implementations29 Jun 2021 Yichi Zhou, Shihong Song, Huishuai Zhang, Jun Zhu, Wei Chen, Tie-Yan Liu

However, it is in general unknown how to deriveefficient and effective EE trade-off methods for non-linearcomplex tasks, suchas contextual bandit with deep neural network as the reward function.

Multi-Armed Bandits

Strategically-timed State-Observation Attacks on Deep Reinforcement Learning Agents

no code implementations ICML Workshop AML 2021 You Qiaoben, Xinning Zhou, Chengyang Ying, Jun Zhu

Deep reinforcement learning (DRL) policies are vulnerable to the adversarial attack on their observations, which may mislead real-world RL agents to catastrophic failures.

Adversarial Attack Continuous Control +2

Towards Safe Reinforcement Learning via Constraining Conditional Value at Risk

no code implementations ICML Workshop AML 2021 Chengyang Ying, Xinning Zhou, Dong Yan, Jun Zhu

Though deep reinforcement learning (DRL) has obtained substantial success, it may encounter catastrophic failures due to the intrinsic uncertainty caused by stochastic policies and environment variability.

Continuous Control reinforcement-learning +2

Accumulative Poisoning Attacks on Real-time Data

1 code implementation NeurIPS 2021 Tianyu Pang, Xiao Yang, Yinpeng Dong, Hang Su, Jun Zhu

Collecting training data from untrusted sources exposes machine learning services to poisoning adversaries, who maliciously manipulate training data to degrade the model accuracy.

Federated Learning

Quasi-Bayesian Dual Instrumental Variable Regression

1 code implementation NeurIPS 2021 Ziyu Wang, Yuhao Zhou, Tongzheng Ren, Jun Zhu

Recent years have witnessed an upsurge of interest in employing flexible machine learning models for instrumental variable (IV) regression, but the development of uncertainty quantification methodology is still lacking.

Bayesian Inference regression

Understanding Softmax Confidence and Uncertainty

no code implementations9 Jun 2021 Tim Pearce, Alexandra Brintrup, Jun Zhu

It is often remarked that neural networks fail to increase their uncertainty when predicting on data far from the training distribution.

OOD Detection

Nonlinear Hawkes Processes in Time-Varying System

no code implementations9 Jun 2021 Feng Zhou, Quyu Kong, Yixuan Zhang, Cheng Feng, Jun Zhu

Hawkes processes are a class of point processes that have the ability to model the self- and mutual-exciting phenomena.

Bayesian Inference Point Processes +1

Stability and Generalization of Bilevel Programming in Hyperparameter Optimization

1 code implementation NeurIPS 2021 Fan Bao, Guoqiang Wu, Chongxuan Li, Jun Zhu, Bo Zhang

Our results can explain some mysterious behaviours of the bilevel programming in practice, for instance, overfitting to the validation set.

Hyperparameter Optimization

Exploring Memorization in Adversarial Training

1 code implementation ICLR 2022 Yinpeng Dong, Ke Xu, Xiao Yang, Tianyu Pang, Zhijie Deng, Hang Su, Jun Zhu

In this paper, we explore the memorization effect in adversarial training (AT) for promoting a deeper understanding of model capacity, convergence, generalization, and especially robust overfitting of the adversarially trained models.

Memorization

Physics-Guided Discovery of Highly Nonlinear Parametric Partial Differential Equations

no code implementations2 Jun 2021 Yingtao Luo, Qiang Liu, Yuntian Chen, WenBo Hu, Jun Zhu

Especially, the discovery of PDEs with highly nonlinear coefficients from low-quality data remains largely under-addressed.

Density Estimation Model Optimization

Two Coupled Rejection Metrics Can Tell Adversarial Examples Apart

1 code implementation CVPR 2022 Tianyu Pang, Huishuai Zhang, Di He, Yinpeng Dong, Hang Su, Wei Chen, Jun Zhu, Tie-Yan Liu

Along with this routine, we find that confidence and a rectified confidence (R-Con) can form two coupled rejection metrics, which could provably distinguish wrongly classified inputs from correctly classified ones.

Unsupervised Part Segmentation through Disentangling Appearance and Shape

no code implementations CVPR 2021 Shilong Liu, Lei Zhang, Xiao Yang, Hang Su, Jun Zhu

We study the problem of unsupervised discovery and segmentation of object parts, which, as an intermediate local representation, are capable of finding intrinsic object structure and providing more explainable recognition results.

Disentanglement Semantic Segmentation

Scalable Quasi-Bayesian Inference for Instrumental Variable Regression

no code implementations NeurIPS 2021 Ziyu Wang, Yuhao Zhou, Tongzheng Ren, Jun Zhu

Recent years have witnessed an upsurge of interest in employing flexible machine learning models for instrumental variable (IV) regression, but the development of uncertainty quantification methodology is still lacking.

Bayesian Inference regression

Rethinking and Reweighting the Univariate Losses for Multi-Label Ranking: Consistency and Generalization

no code implementations NeurIPS 2021 Guoqiang Wu, Chongxuan Li, Kun Xu, Jun Zhu

Our results show that learning algorithms with the consistent univariate loss have an error bound of $O(c)$ ($c$ is the number of labels), while algorithms with the inconsistent pairwise loss depend on $O(\sqrt{c})$ as shown in prior work.

Multi-Label Classification

Automated Decision-based Adversarial Attacks

no code implementations9 May 2021 Qi-An Fu, Yinpeng Dong, Hang Su, Jun Zhu

Deep learning models are vulnerable to adversarial examples, which can fool a target classifier by imposing imperceptible perturbations onto natural examples.

Adversarial Attack Program Synthesis

MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering

1 code implementation ICLR 2021 Tsung Wei Tsai, Chongxuan Li, Jun Zhu

We present Mixture of Contrastive Experts (MiCE), a unified probabilistic clustering framework that simultaneously exploits the discriminative representations learned by contrastive learning and the semantic structures captured by a latent mixture model.

Contrastive Learning Image Clustering

Few-shot Continual Learning: a Brain-inspired Approach

no code implementations19 Apr 2021 Liyuan Wang, Qian Li, Yi Zhong, Jun Zhu

Our solution is based on the observation that continual learning of a task sequence inevitably interferes few-shot generalization, which makes it highly nontrivial to extend few-shot learning strategies to continual learning scenarios.

Continual Learning Few-Shot Learning

Counter-Strike Deathmatch with Large-Scale Behavioural Cloning

3 code implementations9 Apr 2021 Tim Pearce, Jun Zhu

This paper describes an AI agent that plays the popular first-person-shooter (FPS) video game `Counter-Strike; Global Offensive' (CSGO) from pixel input.

Behavioural cloning FPS Games

Accurate and Reliable Forecasting using Stochastic Differential Equations

no code implementations28 Mar 2021 Peng Cui, Zhijie Deng, WenBo Hu, Jun Zhu

It is critical yet challenging for deep learning models to properly characterize uncertainty that is pervasive in real-world environments.

Prediction Intervals

LiBRe: A Practical Bayesian Approach to Adversarial Detection

1 code implementation CVPR 2021 Zhijie Deng, Xiao Yang, Shizhen Xu, Hang Su, Jun Zhu

Despite their appealing flexibility, deep neural networks (DNNs) are vulnerable against adversarial examples.

Adversarial Defense

Black-box Detection of Backdoor Attacks with Limited Information and Data

no code implementations ICCV 2021 Yinpeng Dong, Xiao Yang, Zhijie Deng, Tianyu Pang, Zihao Xiao, Hang Su, Jun Zhu

Although deep neural networks (DNNs) have made rapid progress in recent years, they are vulnerable in adversarial environments.

Implicit Normalizing Flows

1 code implementation ICLR 2021 Cheng Lu, Jianfei Chen, Chongxuan Li, Qiuhao Wang, Jun Zhu

Through theoretical analysis, we show that the function space of ImpFlow is strictly richer than that of ResFlows.

DNN2LR: Automatic Feature Crossing for Credit Scoring

no code implementations24 Feb 2021 Qiang Liu, Zhaocheng Liu, Haoli Zhang, Yuntian Chen, Jun Zhu

Accordingly, we can design an automatic feature crossing method to find feature interactions in DNN, and use them as cross features in LR.

Feature Engineering

Rethinking Natural Adversarial Examples for Classification Models

1 code implementation23 Feb 2021 Xiao Li, Jianmin Li, Ting Dai, Jie Shi, Jun Zhu, Xiaolin Hu

A detection model based on the classification model EfficientNet-B7 achieved a top-1 accuracy of 53. 95%, surpassing previous state-of-the-art classification models trained on ImageNet, suggesting that accurate localization information can significantly boost the performance of classification models on ImageNet-A.

Classification General Classification +2

High-fidelity Prediction of Megapixel Longitudinal Phase-space Images of Electron Beams using Encoder-Decoder Neural Networks

no code implementations25 Jan 2021 Jun Zhu, Ye Chen, Frank Brinker, Winfried Decking, Sergey Tomin, Holger Schlarb

We also show the scalability and interpretability of the model by sharing the same decoder with more than one encoder used for different setups of the photoinjector, and propose a pragmatic way to model a facility with various diagnostics and working points.

Cognitive Visual Inspection Service for LCD Manufacturing Industry

no code implementations11 Jan 2021 Yuanyuan Ding, Junchi Yan, Guoqiang Hu, Jun Zhu

This paper discloses a novel visual inspection system for liquid crystal display (LCD), which is currently a dominant type in the FPD industry.

Defect Detection

Relaxed Conditional Image Transfer for Semi-supervised Domain Adaptation

no code implementations5 Jan 2021 Qijun Luo, Zhili Liu, Lanqing Hong, Chongxuan Li, Kuo Yang, Liyuan Wang, Fengwei Zhou, Guilin Li, Zhenguo Li, Jun Zhu

Semi-supervised domain adaptation (SSDA), which aims to learn models in a partially labeled target domain with the assistance of the fully labeled source domain, attracts increasing attention in recent years.

Domain Adaptation

Adaptive N-step Bootstrapping with Off-policy Data

no code implementations1 Jan 2021 Guan Wang, Dong Yan, Hang Su, Jun Zhu

In this work, we point out that the optimal value of n actually differs on each data point, while the fixed value n is a rough average of them.

Atari Games

Ranking Cost: One-Stage Circuit Routing by Directly Optimizing Global Objective Function

no code implementations1 Jan 2021 Shiyu Huang, Bin Wang, Dong Li, Jianye Hao, Jun Zhu, Ting Chen

In our method, we introduce a new set of variables called cost maps, which can help the A* router to find out proper paths to achieve the global object.

Series Saliency: Temporal Interpretation for Multivariate Time Series Forecasting

no code implementations16 Dec 2020 Qingyi Pan, WenBo Hu, Jun Zhu

Though deep learning methods have recently been developed to give superior forecasting results, it is crucial to improve the interpretability of time series models.

Data Augmentation Multivariate Time Series Forecasting

Fork or Fail: Cycle-Consistent Training with Many-to-One Mappings

1 code implementation14 Dec 2020 Qipeng Guo, Zhijing Jin, Ziyu Wang, Xipeng Qiu, Weinan Zhang, Jun Zhu, Zheng Zhang, David Wipf

Cycle-consistent training is widely used for jointly learning a forward and inverse mapping between two domains of interest without the cumbersome requirement of collecting matched pairs within each domain.

Knowledge Graphs Text Generation

Understanding and Exploring the Network with Stochastic Architectures

1 code implementation NeurIPS 2020 Zhijie Deng, Yinpeng Dong, Shifeng Zhang, Jun Zhu

In this work, we decouple the training of a network with stochastic architectures (NSA) from NAS and provide a first systematical investigation on it as a stand-alone problem.

Neural Architecture Search

Multi-label classification: do Hamming loss and subset accuracy really conflict with each other?

1 code implementation NeurIPS 2020 Guoqiang Wu, Jun Zhu

On the other hand, when directly optimizing SA with its surrogate loss, it has learning guarantees that depend on $O(\sqrt{c})$ for both HL and SA measures.

General Classification Multi-Label Classification

Further Analysis of Outlier Detection with Deep Generative Models

1 code implementation NeurIPS 2020 Ziyu Wang, Bin Dai, David Wipf, Jun Zhu

The recent, counter-intuitive discovery that deep generative models (DGMs) can frequently assign a higher likelihood to outliers has implications for both outlier detection applications as well as our overall understanding of generative modeling.

Outlier Detection

Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models

1 code implementation NeurIPS Workshop ICBINB 2020 Fan Bao, Kun Xu, Chongxuan Li, Lanqing Hong, Jun Zhu, Bo Zhang

The learning and evaluation of energy-based latent variable models (EBLVMs) without any structural assumptions are highly challenging, because the true posteriors and the partition functions in such models are generally intractable.

Bi-level Score Matching for Learning Energy-based Latent Variable Models

1 code implementation NeurIPS 2020 Fan Bao, Chongxuan Li, Kun Xu, Hang Su, Jun Zhu, Bo Zhang

This paper presents a bi-level score matching (BiSM) method to learn EBLVMs with general structures by reformulating SM as a bi-level optimization problem.

Stochastic Optimization

BayesAdapter: Being Bayesian, Inexpensively and Reliably, via Bayesian Fine-tuning

1 code implementation5 Oct 2020 Zhijie Deng, Jun Zhu

Despite their theoretical appealingness, Bayesian neural networks (BNNs) are left behind in real-world adoption, mainly due to persistent concerns on their scalability, accessibility, and reliability.

Variational Inference

Bag of Tricks for Adversarial Training

2 code implementations ICLR 2021 Tianyu Pang, Xiao Yang, Yinpeng Dong, Hang Su, Jun Zhu

Adversarial training (AT) is one of the most effective strategies for promoting model robustness.

Adversarial Robustness

BayesAdapter: Being Bayesian, Inexpensively and Robustly, via Bayesian Fine-tuning

no code implementations28 Sep 2020 Zhijie Deng, Xiao Yang, Hao Zhang, Yinpeng Dong, Jun Zhu

Despite their theoretical appealingness, Bayesian neural networks (BNNs) are falling far behind in terms of adoption in real-world applications compared with normal NNs, mainly due to their limited scalability in training, and low fidelity in their uncertainty estimates.

Variational Inference

Switching Transferable Gradient Directions for Query-Efficient Black-Box Adversarial Attacks

no code implementations15 Sep 2020 Chen Ma, Shuyu Cheng, Li Chen, Jun Zhu, Junhai Yong

In each iteration, SWITCH first tries to update the current sample along the direction of $\hat{\mathbf{g}}$, but considers switching to its opposite direction $-\hat{\mathbf{g}}$ if our algorithm detects that it does not increase the value of the attack objective function.

Adversarial Attack

Training Interpretable Convolutional Neural Networks by Differentiating Class-specific Filters

1 code implementation ECCV 2020 Haoyu Liang, Zhihao Ouyang, Yuyuan Zeng, Hang Su, Zihao He, Shu-Tao Xia, Jun Zhu, Bo Zhang

Most existing works attempt post-hoc interpretation on a pre-trained model, while neglecting to reduce the entanglement underlying the model.

Object Localization

RobFR: Benchmarking Adversarial Robustness on Face Recognition

2 code implementations8 Jul 2020 Xiao Yang, Dingcheng Yang, Yinpeng Dong, Hang Su, Wenjian Yu, Jun Zhu

Based on large-scale evaluations, the commercial FR API services fail to exhibit acceptable performance on robustness evaluation, and we also draw several important conclusions for understanding the adversarial robustness of FR models and providing insights for the design of robust FR models.

Adversarial Robustness Face Recognition

Efficient Learning of Generative Models via Finite-Difference Score Matching

1 code implementation NeurIPS 2020 Tianyu Pang, Kun Xu, Chongxuan Li, Yang song, Stefano Ermon, Jun Zhu

Several machine learning applications involve the optimization of higher-order derivatives (e. g., gradients of gradients) during training, which can be expensive in respect to memory and computation even with automatic differentiation.

Efficient Inference of Flexible Interaction in Spiking-neuron Networks

no code implementations ICLR 2021 Feng Zhou, Yixuan Zhang, Jun Zhu

Hawkes process provides an effective statistical framework for analyzing the time-dependent interaction of neuronal spiking activities.

Dynamic Window-level Granger Causality of Multi-channel Time Series

no code implementations14 Jun 2020 Zhiheng Zhang, Wen-Bo Hu, Tian Tian, Jun Zhu

In this paper, we present the dynamic window-level Granger causality method (DWGC) for multi-channel time series data.

Time Series

Brain-inspired global-local learning incorporated with neuromorphic computing

no code implementations5 Jun 2020 Yujie Wu, Rong Zhao, Jun Zhu, Feng Chen, Mingkun Xu, Guoqi Li, Sen Song, Lei Deng, Guanrui Wang, Hao Zheng, Jing Pei, Youhui Zhang, Mingguo Zhao, Luping Shi

We demonstrate the advantages of this model in multiple different tasks, including few-shot learning, continual learning, and fault-tolerance learning in neuromorphic vision sensors.

Continual Learning Few-Shot Learning

Nonparametric Score Estimators

1 code implementation ICML 2020 Yuhao Zhou, Jiaxin Shi, Jun Zhu

Estimating the score, i. e., the gradient of log density function, from a set of samples generated by an unknown distribution is a fundamental task in inference and learning of probabilistic models that involve flexible yet intractable densities.

Lazy-CFR: fast and near-optimal regret minimization for extensive games with imperfect information

no code implementations ICLR 2020 Yichi Zhou, Tongzheng Ren, Jialian Li, Dong Yan, Jun Zhu

In this paper, we present Lazy-CFR, a CFR algorithm that adopts a lazy update strategy to avoid traversing the whole game tree in each round.

SUMO: Unbiased Estimation of Log Marginal Probability for Latent Variable Models

no code implementations ICLR 2020 Yucen Luo, Alex Beatson, Mohammad Norouzi, Jun Zhu, David Duvenaud, Ryan P. Adams, Ricky T. Q. Chen

Standard variational lower bounds used to train latent variable models produce biased estimates of most quantities of interest.

Towards Face Encryption by Generating Adversarial Identity Masks

1 code implementation ICCV 2021 Xiao Yang, Yinpeng Dong, Tianyu Pang, Hang Su, Jun Zhu, Yuefeng Chen, Hui Xue

As billions of personal data being shared through social media and network, the data privacy and security have drawn an increasing attention.

Face Recognition

Triple Memory Networks: a Brain-Inspired Method for Continual Learning

no code implementations6 Mar 2020 Liyuan Wang, Bo Lei, Qian Li, Hang Su, Jun Zhu, Yi Zhong

Continual acquisition of novel experience without interfering previously learned knowledge, i. e. continual learning, is critical for artificial neural networks, but limited by catastrophic forgetting.

class-incremental learning Hippocampus +1

VFlow: More Expressive Generative Flows with Variational Data Augmentation

1 code implementation ICML 2020 Jianfei Chen, Cheng Lu, Biqi Chenli, Jun Zhu, Tian Tian

Generative flows are promising tractable models for density modeling that define probabilistic distributions with invertible transformations.

Ranked #25 on Image Generation on CIFAR-10 (bits/dimension metric)

Density Estimation Image Generation +2

Boosting Adversarial Training with Hypersphere Embedding

1 code implementation NeurIPS 2020 Tianyu Pang, Xiao Yang, Yinpeng Dong, Kun Xu, Jun Zhu, Hang Su

Adversarial training (AT) is one of the most effective defenses against adversarial attacks for deep learning models.

Representation Learning

A Wasserstein Minimum Velocity Approach to Learning Unnormalized Models

1 code implementation pproximateinference AABI Symposium 2019 Ziyu Wang, Shuyu Cheng, Yueru Li, Jun Zhu, Bo Zhang

Score matching provides an effective approach to learning flexible unnormalized models, but its scalability is limited by the need to evaluate a second-order derivative.

Adversarial Distributional Training for Robust Deep Learning

1 code implementation NeurIPS 2020 Yinpeng Dong, Zhijie Deng, Tianyu Pang, Hang Su, Jun Zhu

Adversarial training (AT) is among the most effective techniques to improve model robustness by augmenting training data with adversarial examples.

Analyzing the Noise Robustness of Deep Neural Networks

no code implementations26 Jan 2020 Kelei Cao, Mengchen Liu, Hang Su, Jing Wu, Jun Zhu, Shixia Liu

The key is to compare and analyze the datapaths of both the adversarial and normal examples.

Adversarial Attack

SVQN: Sequential Variational Soft Q-Learning Networks

no code implementations ICLR 2020 Shiyu Huang, Hang Su, Jun Zhu, Ting Chen

Partially Observable Markov Decision Processes (POMDPs) are popular and flexible models for real-world decision-making applications that demand the information from past observations to make optimal decisions.

Decision Making Q-Learning +2

Benchmarking Adversarial Robustness

no code implementations26 Dec 2019 Yinpeng Dong, Qi-An Fu, Xiao Yang, Tianyu Pang, Hang Su, Zihao Xiao, Jun Zhu

Deep neural networks are vulnerable to adversarial examples, which becomes one of the most important research problems in the development of deep learning.

Adversarial Attack Adversarial Robustness +1

Triple Generative Adversarial Networks

1 code implementation20 Dec 2019 Chongxuan Li, Kun Xu, Jiashuo Liu, Jun Zhu, Bo Zhang

It is formulated as a three-player minimax game consisting of a generator, a classifier and a discriminator, and therefore is referred to as Triple Generative Adversarial Network (Triple-GAN).

Classification Conditional Image Generation +3

The Search for Sparse, Robust Neural Networks

1 code implementation5 Dec 2019 Justin Cosentino, Federico Zaiter, Dan Pei, Jun Zhu

Recent work on deep neural network pruning has shown there exist sparse subnetworks that achieve equal or improved accuracy, training time, and loss using fewer network parameters when compared to their dense counterparts.

Network Pruning

Design and Interpretation of Universal Adversarial Patches in Face Detection

no code implementations ECCV 2020 Xiao Yang, Fangyun Wei, Hongyang Zhang, Jun Zhu

We consider universal adversarial patches for faces -- small visual elements whose addition to a face image reliably destroys the performance of face detectors.

Face Detection

Measuring Uncertainty through Bayesian Learning of Deep Neural Network Structure

1 code implementation22 Nov 2019 Zhijie Deng, Yucen Luo, Jun Zhu, Bo Zhang

Bayesian neural networks (BNNs) augment deep networks with uncertainty quantification by Bayesian treatment of the network weights.

Bayesian Inference Neural Architecture Search +1

Generative Well-intentioned Networks

no code implementations NeurIPS 2019 Justin Cosentino, Jun Zhu

We propose Generative Well-intentioned Networks (GWINs), a novel framework for increasing the accuracy of certainty-based, closed-world classifiers.

Understanding and Stabilizing GANs' Training Dynamics with Control Theory

1 code implementation29 Sep 2019 Kun Xu, Chongxuan Li, Jun Zhu, Bo Zhang

There are existing efforts that model the training dynamics of GANs in the parameter space but the analysis cannot directly motivate practically effective stabilizing methods.

Ranked #32 on Image Generation on CIFAR-10 (Inception score metric)

Image Generation L2 Regularization

Deep Bayesian Structure Networks

1 code implementation25 Sep 2019 Zhijie Deng, Yucen Luo, Jun Zhu, Bo Zhang

Bayesian neural networks (BNNs) introduce uncertainty estimation to deep networks by performing Bayesian inference on network weights.

Bayesian Inference Neural Architecture Search +1

Training Interpretable Convolutional Neural Networks towards Class-specific Filters

no code implementations25 Sep 2019 Haoyu Liang, Zhihao Ouyang, Hang Su, Yuyuan Zeng, Zihao He, Shu-Tao Xia, Jun Zhu, Bo Zhang

Convolutional neural networks (CNNs) have often been treated as “black-box” and successfully used in a range of tasks.

Mixup Inference: Better Exploiting Mixup to Defend Adversarial Attacks

1 code implementation ICLR 2020 Tianyu Pang, Kun Xu, Jun Zhu

Our experiments on CIFAR-10 and CIFAR-100 demonstrate that MI can further improve the adversarial robustness for the models trained by mixup and its variants.

Adversarial Robustness

A Simple yet Effective Baseline for Robust Deep Learning with Noisy Labels

no code implementations20 Sep 2019 Yucen Luo, Jun Zhu, Tomas Pfister

Recently deep neural networks have shown their capacity to memorize training data, even with noisy labels, which hurts generalization performance.

Learning with noisy labels

DashNet: A Hybrid Artificial and Spiking Neural Network for High-speed Object Tracking

no code implementations15 Sep 2019 Zheyu Yang, Yujie Wu, Guanrui Wang, Yukuan Yang, Guoqi Li, Lei Deng, Jun Zhu, Luping Shi

To the best of our knowledge, DashNet is the first framework that can integrate and process ANNs and SNNs in a hybrid paradigm, which provides a novel solution to achieve both effectiveness and efficiency for high-speed object tracking.

Object Tracking

Improving Black-box Adversarial Attacks with a Transfer-based Prior

2 code implementations NeurIPS 2019 Shuyu Cheng, Yinpeng Dong, Tianyu Pang, Hang Su, Jun Zhu

We consider the black-box adversarial setting, where the adversary has to generate adversarial perturbations without access to the target models to compute gradients.

Scalable Training of Inference Networks for Gaussian-Process Models

2 code implementations27 May 2019 Jiaxin Shi, Mohammad Emtiyaz Khan, Jun Zhu

Inference in Gaussian process (GP) models is computationally challenging for large data, and often difficult to approximate with a small number of inducing points.

Rethinking Softmax Cross-Entropy Loss for Adversarial Robustness

2 code implementations ICLR 2020 Tianyu Pang, Kun Xu, Yinpeng Dong, Chao Du, Ning Chen, Jun Zhu

Previous work shows that adversarially robust generalization requires larger sample complexity, and the same dataset, e. g., CIFAR-10, which enables good standard accuracy may not suffice to train robust models.

Adversarial Robustness

Countering Noisy Labels By Learning From Auxiliary Clean Labels

no code implementations23 May 2019 Tsung Wei Tsai, Chongxuan Li, Jun Zhu

We consider the learning from noisy labels (NL) problem which emerges in many real-world applications.

Boosting Generative Models by Leveraging Cascaded Meta-Models

1 code implementation11 May 2019 Fan Bao, Hang Su, Jun Zhu

Besides, our framework can be extended to semi-supervised boosting, where the boosted model learns a joint distribution of data and labels.

$A^*$ sampling with probability matching

no code implementations ICLR 2019 Yichi Zhou, Jun Zhu

We provide insights into the relationship between $A^*$ sampling and probability matching by analyzing a nontrivial special case in which the state space is partitioned into two subsets.

Decision Making

Efficient Decision-based Black-box Adversarial Attacks on Face Recognition

1 code implementation CVPR 2019 Yinpeng Dong, Hang Su, Baoyuan Wu, Zhifeng Li, Wei Liu, Tong Zhang, Jun Zhu

In this paper, we evaluate the robustness of state-of-the-art face recognition models in the decision-based black-box attack setting, where the attackers have no access to the model parameters and gradients, but can only acquire hard-label predictions by sending queries to the target model.

Face Recognition

Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks

1 code implementation CVPR 2019 Yinpeng Dong, Tianyu Pang, Hang Su, Jun Zhu

In this paper, we propose a translation-invariant attack method to generate more transferable adversarial examples against the defense models.

Translation

Cluster Alignment with a Teacher for Unsupervised Domain Adaptation

1 code implementation ICCV 2019 Zhijie Deng, Yucen Luo, Jun Zhu

Deep learning methods have shown promise in unsupervised domain adaptation, which aims to leverage a labeled source domain to learn a classifier for the unlabeled target domain with a different distribution.

Unsupervised Domain Adaptation

Function Space Particle Optimization for Bayesian Neural Networks

1 code implementation ICLR 2019 Ziyu Wang, Tongzheng Ren, Jun Zhu, Bo Zhang

While Bayesian neural networks (BNNs) have drawn increasing attention, their posterior inference remains challenging, due to the high-dimensional and over-parameterized nature.

Variational Inference

Batch Virtual Adversarial Training for Graph Convolutional Networks

no code implementations25 Feb 2019 Zhijie Deng, Yinpeng Dong, Jun Zhu

We present batch virtual adversarial training (BVAT), a novel regularization method for graph convolutional networks (GCNs).

General Classification Node Classification

Understanding MCMC Dynamics as Flows on the Wasserstein Space

1 code implementation1 Feb 2019 Chang Liu, Jingwei Zhuo, Jun Zhu

It is known that the Langevin dynamics used in MCMC is the gradient flow of the KL divergence on the Wasserstein space, which helps convergence analysis and inspires recent particle-based variational inference methods (ParVIs).

Novel Concepts Variational Inference

Reward Shaping via Meta-Learning

no code implementations27 Jan 2019 Haosheng Zou, Tongzheng Ren, Dong Yan, Hang Su, Jun Zhu

Reward shaping is one of the most effective methods to tackle the crucial yet challenging problem of credit assignment in Reinforcement Learning (RL).

Meta-Learning

Improving Adversarial Robustness via Promoting Ensemble Diversity

6 code implementations25 Jan 2019 Tianyu Pang, Kun Xu, Chao Du, Ning Chen, Jun Zhu

Though deep neural networks have achieved significant progress on various tasks, often enhanced by model ensemble, existing high-performance models can be vulnerable to adversarial attacks.

Adversarial Robustness

Towards Interpretable Deep Neural Networks by Leveraging Adversarial Examples

no code implementations25 Jan 2019 Yinpeng Dong, Fan Bao, Hang Su, Jun Zhu

3) We propose to improve the consistency of neurons on adversarial example subset by an adversarial training algorithm with a consistent loss.

To Relieve Your Headache of Training an MRF, Take AdVIL

no code implementations ICLR 2020 Chongxuan Li, Chao Du, Kun Xu, Max Welling, Jun Zhu, Bo Zhang

We propose a black-box algorithm called {\it Adversarial Variational Inference and Learning} (AdVIL) to perform inference and learning on a general Markov random field (MRF).

Variational Inference

Stochastic Expectation Maximization with Variance Reduction

no code implementations NeurIPS 2018 Jianfei Chen, Jun Zhu, Yee Whye Teh, Tong Zhang

However, sEM has a slower asymptotic convergence rate than batch EM, and requires a decreasing sequence of step sizes, which is difficult to tune.