YIFEI MA

马逸飞

Last update: June 16, 2016

Office: 8223 Gates Building
Mails please send to: 6105 Gates Bldg.
5000 Forbes Avenue
Pittsburgh, PA 15213

Email: yifeim at cs.cmu.edu
[Google Calendar]


I am in my fifth year of PhD program in Machine Learning, School of Computer Science, Carnegie Mellon University. I seek to develop tools that can aid knowledge exploration as well as decision-making in scientific, social, and commercial environments. Viewing machine learning as an approach to find signals in data, one important research is on better human/mobile sensor interaction. Currently, interactive machine learning faces challenges in handling complex conditions and patterns in real applications. My focus is to increase model complexity for better automation in signal searches.

Active search is an example of interactive machine learning. It aims to discover all positive targets by examining as few candidate instances and learning from their outcomes as possible in a sequential and adaptive manner. To apply active search to real and more complex scenarios, I analyzed different assumptions about how these instances are organized, how to access them, and how to define rewards.

The methods that I consider generally belong to the families of Bayesian optimization and multi-armed bandits. For the theory part, I am interested in regret analysis, spectral graph theories, as well as diversity via submodular set functions. I also consider information/statistical theory to demonstrate optimality despite the adaptive and uncertain nature of active systems. My applications are in robotics and information retrieval. In the future, I want to also consider complexity from partial observability, system dynamics, and outliers, for applications with deep and dynamic natures.

[Resume]




Publications

slides

Conferences

  1. Yifei Ma, Tzu-Kuo Huang, Jeff Schneider. Active Search and Bandits on Graphs Using Sigma-Optimality. UAI 2015. [paper] [codes] [spotlight] [poster]
  2. Yifei Ma, Dougal J. Sutherland, Roman Garnett, Jeff Schneider. Active Pointillistic Pattern Search. AISTATS 2015. Two Shared Lead Authors. [paper] [supp] [spotlight] [poster(pptx)]
  3. Yifei Ma, Roman Garnett, Jeff Schneider. Active Area Search via Bayesian Quadrature. AISTATS 2014. [paper]
  4. Yifei Ma, Roman Garnett, Jeff Schneider. Sigma-Optimality for Active Learning on Gaussian Random Fields. NIPS 2013. [paper] [codes] [poster]
  5. Guangyu Xia, Tongbo Huang, Yifei Ma, Roger B. Dannenberg, Christos Faloutsos. MidiFind: Similarity Search and Popularity Mining in Large MIDI Databases. CMMR 2013: 259-276.
  6. Yifei Ma, Li Li, Xiaolin Huang, Shuning Wang, Robust Support Vector Machine Using Least Median Loss Penalty, Proceedings of the 18th IFAC World Congress, Volume 18, Part 1, 2011. [Submitted pdf][Details]

Workshops

  1. Yifei Ma, Roman Garnett, Jeff Schneider. Active Search for Sparse Signals with Region Sensing. ICML 2016 Workshop on Data-Efficient Machine Learning. [paper] [workshop]
  2. Yifei Ma, Dougal J. Sutherland, Roman Garnett, Jeff Schneider. Active Pointillistic Pattern Search. NIPS 2014 Workshop on Bayesian Optimization.
  3. Yifei Ma, Roman Garnett, Jeff Schneider. Submodularity in Batch Active Learning and Survey Problems on Gaussian Random Fields. NIPS 2012 Workshop on DISCML.

Experience


Citadel LLC. Quantitative Researcher Intern. 5/2014-8/2014 at Chicago.


Education

Ph.D. Student in Machine Learning, Carnegie Mellon University, 8/2011-now.

B.S. in Automation, Dual B.S. in Mathematics, Tsinghua University, 8/2007-7/2011.

Exchange Study (Credits Transferred), Georgia Inst. of Technology, 8/2009-12/2009.

Honors and Awards

Volunteer Experience

Student Activities and Hobbies

Photo of me.