Search Results for author: Zhiqiang Xu

Found 41 papers, 16 papers with code

Finite Samples for Shallow Neural Networks

no code implementations17 Mar 2025 Yu Xia, Zhiqiang Xu

This paper investigates the ability of finite samples to identify two-layer irreducible shallow networks with various nonlinear activation functions, including rectified linear units (ReLU) and analytic functions such as the logistic sigmoid and hyperbolic tangent.

MagicDistillation: Weak-to-Strong Video Distillation for Large-Scale Few-Step Synthesis

no code implementations17 Mar 2025 Shitong Shao, Hongwei Yi, Hanzhong Guo, Tian Ye, Daquan Zhou, Michael Lingelbach, Zhiqiang Xu, Zeke Xie

To address these challenges, we propose MagicDistillation, a novel framework designed to reduce inference overhead while ensuring the generalization of VDMs for portrait video synthesis.

Pastiche Novel Generation Creating: Fan Fiction You Love in Your Favorite Author's Style

no code implementations21 Feb 2025 Xueran Han, YuHan Liu, Mingzhe Li, Wei Liu, Sen Hu, Rui Yan, Zhiqiang Xu, Xiuying Chen

Great novels create immersive worlds with rich character arcs, well-structured plots, and nuanced writing styles.

FedMHO: Heterogeneous One-Shot Federated Learning Towards Resource-Constrained Edge Devices

1 code implementation12 Feb 2025 Dezhong Yao, Yuexin Shi, Tongtong Liu, Zhiqiang Xu

One-shot FL has emerged as a promising approach to mitigate communication overhead, and model-heterogeneous FL solves the problem of diverse computing resources across clients.

Edge-computing Federated Learning

Principled Data Selection for Alignment: The Hidden Risks of Difficult Examples

1 code implementation11 Feb 2025 Chengqian Gao, Haonan Li, Liu Liu, Zeke Xie, Peilin Zhao, Zhiqiang Xu

Challenging this, we propose a new principle: Preference data vary in difficulty, and overly difficult examples hinder alignment, by exceeding the model's capacity.

A Simple and Efficient Baseline for Zero-Shot Generative Classification

no code implementations17 Dec 2024 Zipeng Qi, Buhua Liu, Shiyan Zhang, Bao Li, Zhiqiang Xu, Haoyi Xiong, Zeke Xie

While recent zero-shot diffusion-based classifiers have made performance advancement on benchmark datasets, they still suffered badly from extremely slow classification speed (e. g., ~1000 seconds per classifying single image on ImageNet).

zero-shot-classification Zero-Shot Learning

Zigzag Diffusion Sampling: Diffusion Models Can Self-Improve via Self-Reflection

1 code implementation14 Dec 2024 Lichen Bai, Shitong Shao, Zikai Zhou, Zipeng Qi, Zhiqiang Xu, Haoyi Xiong, Zeke Xie

Diffusion models, the most popular generative paradigm so far, can inject conditional information into the generation path to guide the latent towards desired directions.

Denoising

Exploring the Generalization Capabilities of AID-based Bi-level Optimization

no code implementations25 Nov 2024 Congliang Chen, Li Shen, Zhiqiang Xu, Wei Liu, Zhi-Quan Luo, Peilin Zhao

We conduct a convergence analysis for a carefully chosen step size to maintain stability.

Bag of Design Choices for Inference of High-Resolution Masked Generative Transformer

2 code implementations16 Nov 2024 Shitong Shao, Zikai Zhou, Tian Ye, Lichen Bai, Zhiqiang Xu, Zeke Xie

Text-to-image diffusion models (DMs) develop at an unprecedented pace, supported by thorough theoretical exploration and empirical analysis.

Text Generation

Golden Noise for Diffusion Models: A Learning Framework

1 code implementation14 Nov 2024 Zikai Zhou, Shitong Shao, Lichen Bai, Zhiqiang Xu, Bo Han, Zeke Xie

With the prepared NPD as the training dataset, we trained a small \textit{noise prompt network}~(NPNet) that can directly learn to transform a random noise into a golden noise.

Prompt Learning

Online Parallel Multi-Task Relationship Learning via Alternating Direction Method of Multipliers

1 code implementation9 Nov 2024 Ruiyu Li, Peilin Zhao, Guangxia Li, Zhiqiang Xu, XueWei Li

In a classical distributed computing architecture with a central server, the proposed OMTL algorithm with the ADMM optimizer outperforms SGD-based approaches in terms of accuracy and efficiency.

Distributed Computing Multi-Task Learning

On the Comparison between Multi-modal and Single-modal Contrastive Learning

no code implementations5 Nov 2024 Wei Huang, Andi Han, Yongqiang Chen, Yuan Cao, Zhiqiang Xu, Taiji Suzuki

Our analysis provides a unified framework that can characterize the optimization and generalization of both single-modal and multi-modal contrastive learning.

Contrastive Learning Learning Theory

Corrected Soft Actor Critic for Continuous Control

no code implementations22 Oct 2024 Yanjun Chen, Xinming Zhang, Xianghui Wang, Zhiqiang Xu, Xiaoyu Shen, Wei zhang

The Soft Actor-Critic (SAC) algorithm is known for its stability and high sample efficiency in deep reinforcement learning.

continuous-control Continuous Control +1

TextLap: Customizing Language Models for Text-to-Layout Planning

1 code implementation9 Oct 2024 Jian Chen, Ruiyi Zhang, Yufan Zhou, Jennifer Healey, Jiuxiang Gu, Zhiqiang Xu, Changyou Chen

Automatic generation of graphical layouts is crucial for many real-world applications, including designing posters, flyers, advertisements, and graphical user interfaces.

Image Generation Natural Language Understanding

Visual Question Decomposition on Multimodal Large Language Models

no code implementations28 Sep 2024 Haowei Zhang, Jianzhe Liu, Zhen Han, Shuo Chen, Bailan He, Volker Tresp, Zhiqiang Xu, Jindong Gu

The finetuning pipeline consists of our proposed dataset and a training objective for selective decomposition.

Visual Question Answering (VQA)

Alignment of Diffusion Models: Fundamentals, Challenges, and Future

1 code implementation11 Sep 2024 Buhua Liu, Shitong Shao, Bao Li, Lichen Bai, Zhiqiang Xu, Haoyi Xiong, James Kwok, Sumi Helal, Zeke Xie

Diffusion models have emerged as the leading paradigm in generative modeling, excelling in various applications.

DALD: Improving Logits-based Detector without Logits from Black-box LLMs

1 code implementation7 Jun 2024 Cong Zeng, Shengkun Tang, Xianjun Yang, Yuanzhou Chen, Yiyou Sun, Zhiqiang Xu, Yao Li, Haifeng Chen, Wei Cheng, Dongkuan Xu

However, these methods grapple with the misalignment between the distributions of the surrogate and the often undisclosed target models, leading to performance degradation, particularly with the introduction of new, closed-source models.

Text Detection Text Generation

Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples

no code implementations6 Jun 2024 Dake Bu, Wei Huang, Taiji Suzuki, Ji Cheng, Qingfu Zhang, Zhiqiang Xu, Hau-San Wong

We further prove that this shared principle is the key to their success-achieve small test error within a small labeled set.

Active Learning Diversity

MGCP: A Multi-Grained Correlation based Prediction Network for Multivariate Time Series

no code implementations30 May 2024 Zhicheng Chen, Xi Xiao, Ke Xu, Zhong Zhang, Yu Rong, Qing Li, Guojun Gan, Zhiqiang Xu, Peilin Zhao

Multivariate time series prediction is widely used in daily life, which poses significant challenges due to the complex correlations that exist at multi-grained levels.

Prediction Time Series +1

Variational Bayes for Federated Continual Learning

1 code implementation23 May 2024 Dezhong Yao, Sanmu Li, Yutong Dai, Zhiqiang Xu, Shengshan Hu, Peilin Zhao, Lichao Sun

Federated continual learning (FCL) has received increasing attention due to its potential in handling real-world streaming data, characterized by evolving data distributions and varying client classes over time.

Continual Learning Federated Learning

Hard-Thresholding Meets Evolution Strategies in Reinforcement Learning

1 code implementation2 May 2024 Chengqian Gao, William de Vazelhes, Hualin Zhang, Bin Gu, Zhiqiang Xu

Evolution Strategies (ES) have emerged as a competitive alternative for model-free reinforcement learning, showcasing exemplary performance in tasks like Mujoco and Atari.

Decision Making MuJoCo +2

Towards Aligned Layout Generation via Diffusion Model with Aesthetic Constraints

1 code implementation7 Feb 2024 Jian Chen, Ruiyi Zhang, Yufan Zhou, Rajiv Jain, Zhiqiang Xu, Ryan Rossi, Changyou Chen

Controllable layout generation refers to the process of creating a plausible visual arrangement of elements within a graphic design (e. g., document and web designs) with constraints representing design intentions.

Layout Design Layout Generation

Behavior Optimized Image Generation

no code implementations18 Nov 2023 Varun Khurana, Yaman K Singla, Jayakumar Subramanian, Rajiv Ratn Shah, Changyou Chen, Zhiqiang Xu, Balaji Krishnamurthy

We show that BoigLLM outperforms 13x larger models such as GPT-3. 5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world.

Image Generation Marketing

SR-R$^2$KAC: Improving Single Image Defocus Deblurring

no code implementations30 Jul 2023 Peng Tang, Zhiqiang Xu, Pengfei Wei, Xiaobin Hu, Peilin Zhao, Xin Cao, Chunlai Zhou, Tobias Lasser

To further alleviate the contingent effect of recursive stacking, i. e., ringing artifacts, we add identity shortcuts between atrous convolutions to simulate residual deconvolutions.

Deblurring Image Defocus Deblurring

A Cover Time Study of a non-Markovian Algorithm

no code implementations8 Jun 2023 Guanhua Fang, Gennady Samorodnitsky, Zhiqiang Xu

In this work, we stand on a theoretical perspective and show that the negative feedback strategy (a count-based exploration method) is better than the naive random walk search.

Label-Retrieval-Augmented Diffusion Models for Learning from Noisy Labels

1 code implementation NeurIPS 2023 Jian Chen, Ruiyi Zhang, Tong Yu, Rohan Sharma, Zhiqiang Xu, Tong Sun, Changyou Chen

Remarkably, by incorporating conditional information from the powerful CLIP model, our method can boost the current SOTA accuracy by 10-20 absolute points in many cases.

 Ranked #1 on Image Classification on Food-101N (using extra training data)

Image Classification Retrieval

Robust Offline Reinforcement Learning with Gradient Penalty and Constraint Relaxation

1 code implementation19 Oct 2022 Chengqian Gao, Ke Xu, Liu Liu, Deheng Ye, Peilin Zhao, Zhiqiang Xu

A promising paradigm for offline reinforcement learning (RL) is to constrain the learned policy to stay close to the dataset behaviors, known as policy constraint offline RL.

D4RL MuJoCo +4

No existence of linear algorithm for Fourier phase retrieval

no code implementations13 Sep 2022 Meng Huang, Zhiqiang Xu

Fourier phase retrieval, which seeks to reconstruct a signal from its Fourier magnitude, is of fundamental importance in fields of engineering and science.

Retrieval

Unsupervised Video Domain Adaptation for Action Recognition: A Disentanglement Perspective

1 code implementation NeurIPS 2023 Pengfei Wei, Lingdong Kong, Xinghua Qu, Yi Ren, Zhiqiang Xu, Jing Jiang, Xiang Yin

Specifically, we consider the generation of cross-domain videos from two sets of latent factors, one encoding the static information and another encoding the dynamic information.

Action Recognition Disentanglement +1

A Comprehensively Tight Analysis of Gradient Descent for PCA

no code implementations NeurIPS 2021 Zhiqiang Xu, Ping Li

We further give the first worst-case analysis that achieves a rate of convergence at $O(\frac{1}{\epsilon}\log\frac{1}{\epsilon})$.

Fairness-aware Federated Learning

no code implementations29 Sep 2021 Zhuozhuo Tu, Zhiqiang Xu, Tairan Huang, DaCheng Tao, Ping Li

Federated Learning is a machine learning technique where a network of clients collaborates with a server to learn a centralized model while keeping data localized.

Fairness Federated Learning +1

Towards Better Generalization of Adaptive Gradient Methods

no code implementations NeurIPS 2020 Yingxue Zhou, Belhal Karimi, Jinxing Yu, Zhiqiang Xu, Ping Li

Adaptive gradient methods such as AdaGrad, RMSprop and Adam have been optimizers of choice for deep learning due to their fast training speed.

On the Overlooked Pitfalls of Weight Decay and How to Mitigate Them: A Gradient-Norm Perspective

1 code implementation NeurIPS 2023 Zeke Xie, Zhiqiang Xu, Jingzhao Zhang, Issei Sato, Masashi Sugiyama

Weight decay is a simple yet powerful regularization technique that has been very widely used in training of deep neural networks (DNNs).

Towards Practical Alternating Least-Squares for CCA

no code implementations NeurIPS 2019 Zhiqiang Xu, Ping Li

To promote the practical use of ALS for CCA, we propose truly alternating least-squares.

Gradient Descent Meets Shift-and-Invert Preconditioning for Eigenvector Computation

no code implementations NeurIPS 2018 Zhiqiang Xu

Shift-and-invert preconditioning, as a classic acceleration technique for the leading eigenvector computation, has received much attention again recently, owing to fast least-squares solvers for efficiently approximating matrix inversions in power iterations.

Stochastic Variance Reduced Riemannian Eigensolver

no code implementations26 May 2016 Zhiqiang Xu, Yiping Ke

We generalize it to Riemannian manifolds and realize it to solve the non-convex eigen-decomposition problem.

The performance of orthogonal multi-matching pursuit under RIP

no code implementations19 Oct 2012 Zhiqiang Xu

In particular, for $M=s^a$ with $a\in [0, 1/2]$, OMMP(M) can recover slowly-decaying $s$-sparse signal within $O(s^{1-a})$ iterations.

Cannot find the paper you are looking for? You can Submit a new open access paper.