Get our free extension to see links to code for papers anywhere online!

Add to Chrome

Add to Firefox

Picture for Chenwei Wu

Chenwei Wu

Hiding Data Helps: On the Benefits of Masking for Sparse Coding

Feb 24, 2023
Muthu Chidambaram, Chenwei Wu, Yu Cheng, Rong Ge

Figure 1 for Hiding Data Helps: On the Benefits of Masking for Sparse Coding

Figure 2 for Hiding Data Helps: On the Benefits of Masking for Sparse Coding

Figure 3 for Hiding Data Helps: On the Benefits of Masking for Sparse Coding

* 21 pages, 3 figures

Access Paper or Ask Questions

Towards Understanding the Data Dependency of Mixup-style Training

Oct 14, 2021
Muthu Chidambaram, Xiang Wang, Yuzheng Hu, Chenwei Wu, Rong Ge

Figure 1 for Towards Understanding the Data Dependency of Mixup-style Training

Figure 2 for Towards Understanding the Data Dependency of Mixup-style Training

Figure 3 for Towards Understanding the Data Dependency of Mixup-style Training

Figure 4 for Towards Understanding the Data Dependency of Mixup-style Training

* 25 pages, 13 figures

Access Paper or Ask Questions

Beyond Lazy Training for Over-parameterized Tensor Decomposition

Oct 22, 2020
Xiang Wang, Chenwei Wu, Jason D. Lee, Tengyu Ma, Rong Ge

Figure 1 for Beyond Lazy Training for Over-parameterized Tensor Decomposition

* NeurIPS 2020; the first two authors contribute equally

Access Paper or Ask Questions

Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks

Oct 08, 2020
Yikai Wu, Xingyu Zhu, Chenwei Wu, Annie Wang, Rong Ge

Figure 1 for Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks

Figure 2 for Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks

Figure 3 for Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks

Figure 4 for Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks

* 29 pages, 26 figures. Main text: 8 pages, 6 figures. First two authors have equal contribution and are in alphabetical order

Access Paper or Ask Questions