Welcome

Welcome to my world on the web, a window through which I'd like to share with you my thoughts, my work, and my life.

Dahua Lin

Research Assistant Professor, Toyota Technological Institute at Chicago (Curriculum Vitae)

I am currently a Research Assistant Professor at Toyota Technological Institute at Chicago (TTIC).TTIC is a philanthropically endowed academic computer science institute, dedicated to basic research and graduate education in computer science. It has been very active at the frontier of machine learning and computer vision research.

Before joining TTI Chicago, I was a Ph.D. student at MIT. I was working at the Computer Science and Artificial Intelligence Laboratory (CSAIL), advised by Dr. John Fisher and Professor Eric Grimson. I also worked closely with Professor Alan Willsky. Prior to my adventure in MIT, I received an M.Phil. at the Chinese University of Hong Kong, and a B.Eng. at the University of Science and Technology of China.

During my life as a student, I was fortunate to have three internships at Microsoft Research, respectively at different locations: Silicon Valley (2010), Redmond (2009), and Beijing (2004).

Research Interests

My research revolves around computer vision -- a fascinating field filled with challenging questions. I also work on machine learning, with an aim to develop effective tools to help answering questions arising from vision problems. My current research mainly follows three themes:

  • New machine learning techniques for Big Data

    The explosive growth of data has presented great challenges -- traditional machine learning techniques that were mostly developed over closed datasets of moderate size are difficult to meet the increasing computational demand of handling large scale and dynamically changing data. New techniques, especially those that leverage the latest advancement in nonparametric analysis, stochastic optimization, and parallel computing, are needed to address these challenges.

  • Bayesian nonparametrics for Big Data Analysis

    Nonparametric models are those whose structure can grow during learning, which provide a flexible framework to describe complex data that evolve over time. Despite their theoretical advantages, application of nonparametric models in large-scale analysis are still impeded by two significant difficulties: lack of efficient estimation methods and the technical complexity in modeling complex relations. My work in this domain mainly aims at overcoming these difficulties and thus expanding the application boundary of nonparametric analysis.

  • Data-driven Scene Understanding

    Natural scenes usually present diverse and complex structures. Whereas there has been significant progress in scene understanding, current approaches are still quite limited -- most of them can only handle scenes with certain structures. The sheer amount of images available on the internet provides a great opportunity to move beyond such limits. My current work in the computer vision domain focuses on leveraging large scale datasets and online data to improve scene understanding performance.

For details, please refer to my Research Page and my Publication List.

Awards

It is my great honor to have my work recognized by other researchers working on the same field. Here are some awards I won in the past several years.

  • Outstanding reviewer award at ICCV 2011
  • Best student paper award at NIPS 2010
  • Outstanding reviewer award at ICCV 2009
  • Best master thesis awards of the Engineering school, CUHK 2006
Based on a template from ThemeMakers