Research of Professor Yi Ma

Areas of Interest

My main research interest is in finding the most pertinent mathematical principles for analyzing and understanding high-dimensional sensorial data such as images, so that machines and human can make more intelligent decisions. The research endeavor has led me touch upon the following areas.

Projects

• Hong Kong University: Integrated Autonomous Intelligent Systems.
• Berkeley: Whitebox Transformers via Sparse Rate Reduction (with Yaodong Yu et. al.).
• Berkeley: Deep Networks from the Principle of Rate Reduction (with Yaodong Yu, Ryan Chan, Haozhi Qi, and Chong You).
• Berkeley: General In-hand Object Rotation with Vision and Touch (with Haozhi Qi, Brent Yi, Jitendra Malik, etc.)
• Berkeley: Dexterous Robot Hand Manipulation (with Haozhi Qi, Roberto Calandra, and Jitendra Malik).
• Berkeley: Canonical Factors for Hybrid Neural Fields (with Brent Yi, Weijia Zeng, and Sam Buchanan).
• Berkeley: Pursuit of Large-Scale 3D Structures and Geometry (with Yichao Zhou and Xili Dai).
• UIUC/MSRA: Low-Rank Matrix Recovery via Convex Optimization (with Wright, Ganesh, Lin and Candes et. al.).
• UIUC: Face Recognition via Sparse Representation (with Wright, Ganesh, Yang, Zhou and Wagner et. al.).
• UIUC: Clustering and Classification via Lossy Compression (with Wright Yang, Mobahi, and Rao et. al.).
• UIUC: Generalized Principal Component Analysis (with Huang, Vidal, and Sastry).
• UIUC/Berkeley: Multiple-View Geometry and 3D Reconstruction (with Kosecka, Soatto, and Sastry).

Funded Projects

Federal

• Mathematical and Scientific Foundations of Deep Learning (sponsored by NSF and Simons Foundation).
• Leveraging Egocentric and Allocentric Representations for Navigation (LEARN) (sponored by ONR).
• A Holistic Approach to Reconstructing Urban Scenes (sponsored by NSF-IIS).
• Scene and Task Aware Knowledge Enhanced Compressive Sensing, DARPA KECoM.
• Advances in the Theory and Practice of Low-Rank Matrix Recovery and Modeling (with Emmanuel Candes, sponsored by NSF-CIF).
• Harnessing Sparse and Low-Dimensional Structures for Robust Clustering of Visual Data (sponsored by ONR).
• SGER: Explorations of Robust Image Classification (sponsored by NSF-IIS).
• PHYSNET: Physical Interaction Using the Internet (sponsored by NSF-HCC).
• Control and Sensing under Limited Information (sponsored by NSF-ECS).
• Estimation of Hybrid Models as Algebraic Sets (sponsored by NSF-CCF).
• Estimation of Hybrid Models in Computer Vision (sponsored by ONR).
• Identification of Hybrid Dynamical Systems (sponsored by NSF-EHS).
• Identifying Spatial and Dynamical Patterns from Images (sponsored by NSF-IIS).

University

• HKUST and Berkeley Joint Research Project.
• Tsinghua-Berkeley Shenzhen Institute Research Fund.
• UC Berkeley Faculty Startup Fund.
• ShanghaiTech: Data Science Research Center Fund.
• DARPA Urban Challenge (sponsored by the University of Illinois and others).
• Geometry and Algebra in Computer Vision, RAP Program (sponsored by Math Department, the University of Illinois).
• Vision-Based Control (sponsored by the University of Illinois Research Board).

Industry

• Sony Inc.
• Bytedance Inc.
• Berkeley FHL VIVE Center, HTC/VIVE.
• Adobe Research Sponsorship.
• Microsoft Research Fellowship (sponsored by the Microsoft Live Lab.).

Reflections on Research

• An essay reflecting on the methodology for research: Warning Signs of Bogus Research Progresses.
• Powerpoint slides from a talk on How to do Research.
• Powerpoint slides from a talk on How to Give a Good Presentation.
• Powerpoint slides from my talk at CCF Elite Forum 2014 on Research as a Career.