Zlin’s Extraordinary Research Oasis

Zeal, Excellence, Reliability and Openness

Welcome to the ZERO Lab, the research group lead by Prof. Zhouchen Lin (Zlin), affiliated to School of Artificial Intelligence, Peking University. We research on machine learning and computer vision.

Recruiting

  • Our group at Peking University is recruiting tenure-track faculties and PostDocs (academic or industrial). For the postdoc program in 2020, please refer to the Chinese version or the English version.
  • Our group at Peking University is recruiting Ph.D.s who have strong mathematical abilities (however, this does not imply that you have to come from mathematics department) and great interest in theoretical analysis in order to enjoy with me how to use mathematics to solve real problems elegantly.

Topics

Acceleration ADMM Adversarial Robustness Adversarial Transferability Alternating direction method Alternating Direction Method of Multipliers Bayes error Boosting Color Filter Array Compressed Sensing Compressive Phase Retrieval, computer vision Contextual distance Convergence Analysis Convex Optimization Data Compression Deep Learning Demosaicking Denoising Dictionary Learning Dimensionality reduction Discriminant analysis Document analysis Double quantization Expectation Maximization Face recognition Feature detection Feature extraction forgery Geometric Optimization Handwriting recognition Image Annotation Image classification Image Denoising Image Processing Image Reconstruction Image rectification Image restoration Image Retrieval Image segmentation Laplace equations Learning-based PDEs Light field Linear discriminant analysis Lorentzian geometry Low Rank Low Rank Representation Low-level vision Lumigraph Machine Learning Majorization Minimization Manifold learning Manifolds Matrix Completion matrix decomposition Metric learning Mismatch Removal Neural Network Neural networks Nonconvex Optimization Optimal control Optimization Partial Differential Equation plenoptic functions Pose Estimation Principal component analysis Robust PCA Robust Principal Component Analysis sampling Semantic segmentation Semi-supervised learning Singular value decomposition Sparse Coding Sparse matrices Sparse Representation Spectral clustering Subspace clustering Subspace Recovery Super-Resolution Superresolution

Latest Publications

Symmetry Discovery for Different Data Types. NN, 2025.

Equivariant neural networks incorporate symmetries into their architecture, achieving higher generalization performance. However, …

Explicit Discovery of Nonlinear Symmetries from Dynamic Data. ICML, 2025.

Symmetry is widely applied in problems such as the design of equivariant networks and the discovery of governing equations, but in …

Low-Dimension-to-High-Dimension Generalization And Its Implications for Length Generalization. ICML, 2025.

Low-Dimension-to-High-Dimension (LDHD) generalization, a subset of Out-of-Distribution (OOD) generalization, involves training on a …

Affine Steerable Equivariant Layer for Canonicalization of Neural Networks. ICLR, 2025.

In the field of equivariant networks, achieving affine equivariance, particularly for general group representations, has long been a …

Relational Learning in Pre-Trained Models: A Theory from Hypergraph Recovery Perspective. ICML, 2024.

Foundation Models (FMs) have demonstrated remarkable insights into the relational dynamics of the world, leading to the crucial …

Affine Equivariant Networks Based on Differential Invariants. CVPR, 2024.

Convolutional neural networks benefit from translation equivariance, achieving tremendous success. Equivariant networks further extend …

Task-Robust Pre-Training for Worst-Case Downstream Adaptation. NeurIPS, 2023.

Pre-training has achieved remarkable success when transferred to downstream tasks. In machine learning, we care about not only the good …

Neural ePDOs: Spatially Adaptive Equivariant Partial Differential Operator Based Networks. ICLR, 2023.

Endowing deep learning models with symmetry priors can lead to a considerable performance improvement. As an interesting bridge between …

Contact

  • Wechat account