Titlebar
Site directory
Aaron D'Souza

Bio:
Aaron D'Souza completed his B.E. in Computer Engineering in 1998 from the Thadomal Shahani Engineering College of Mumbai University, India. He joined the Computer Science department at USC as a Masters student in Fall 1998. From Fall 1998 to Spring 2000, he worked as a graduate research assistant at the Intelligent Systems Division of the USC Information Sciences Institute under the supervision of Dr. Lewis Johnson.

In Spring 2000 he joined the Ph.D. program in Computer Science, and since Fall 2000 has been a graduate research assistant at the Computational Learning and Motor Control Lab at USC, under the supervision of Dr. Stefan Schaal.

In October 2004 he successfully defended his dissertation entitled Towards Tractable Parameter-Free Statistical Learning.

Research Interests:
Machine Learning, specifically the automatic determination of model complexity in Statistical Learning through Bayesian inference, with an emphasis on high-dimensional learning applications such as Humanoid Robot Control.

Address:
Aaron D'Souza
Google Inc.
1600 Amphitheatre Parkway,
Mountain View, CA 94043, USA

Publications:

  • J. Ting, A. D'Souza, K. Yamamoto, T. Yoshioka, D. Hoffman, S. Kakei, L. Sergio, J. Kalaska, M. Kawato, P. Strick and S. Schaal. Predicting EMG Data from M1 Neurons with Variational Bayesian Least Squares. In Advances in Neural Information Processing Systems 18 (NIPS 2005) (to appear).
  • Sethu Vijayakumar, Aaron D'Souza and Stefan Schaal. Locally weighted projection regression. Neural Computation. (in press).
  • Jo-Anne Ting, Aaron D'Souza and Stefan Schaal. Predicting EMG Activity from Neural Firing in M1 with Bayesian Backfitting. In Proceedings of the 11th Joint Symposium on Neural Computation (JSNC 2004). Los Angeles, CA. May 2004.
  • Aaron D'Souza. Towards Tractable Parameter-Free Statistical Learning. Ph.D. Thesis, Department of Computer Science, University of Southern California, Los Angeles, CA. 2004.
    [pdf]
  • Aaron D'Souza, Sethu Vijayakumar and Stefan Schaal. The Bayesian Backfitting Relevance Vector Machine. In Proceedings of the International Conference on Machine Learning (ICML 2004).
    [pdf], [ps.gz], poster [pdf]
  • Aaron D'Souza, Sethu Vijayakumar and Stefan Schaal. Bayesian Backfitting. In Proceedings of the 10th Joint Symposium on Neural Computation (JSNC 2003). Irvine, CA. May 2003
    paper [pdf], poster [pdf]
  • Sethu Vijayakumar, Aaron D'Souza, Tomohiro Shibata, Jörg Conradt and Stefan Schaal. Statistical learning for humanoid robots. Autonomous Robots, 12, 55-69, 2002
    [pdf], [ps.gz]
  • Stefan Schaal, Sethu Vijayakumar, Aaron D'Souza, Auke Ijspeert & Jun Nakanishi. Real-time statistical learning for robotics and human augmentation. International Symposium of Robotics Research. Lorne, Victoria, Australia. Nov 2001. Springer (In press)
    [pdf]
  • Aaron D'Souza, Sethu Vijayakumar and Stefan Schaal. Are Internal Models of the Entire Body Learnable? In Society for Neuroscience Abstracts. Vol. 27, Program No. 406.2, 2001.
    (abstract) [pdf], [ps.gz]
  • Aaron D'Souza, Sethu Vijayakumar, and Stefan Schaal. Learning inverse kinematics. In Proceedings of the IEEE International Conference on Intelligent Robots and Systems (IROS 2001), Maui, HI, USA, October 2001.
    [pdf], [ps.gz]
  • Aaron D'Souza, Jeff Rickel, Bruno Herreros, and W. Lewis Johnson. An automated lab instructor for simulated science experiments. In Proceedings of the International Conference on Artificial Intelligence in Education (AI-ED 2001), pp. 65-76, San Antonio, TX, May 2001.
    (Recieved Distinguished Paper Award)
    [pdf], [ps.gz]

Notes to myself (and other students):

  • Bayesian Factor Analysis
    [pdf], [ps.gz]
  • Derivation of the Cramér-Rao bound for a single continous variable.
    [pdf], [ps.gz]
  • Probabilistic Derivation of the basic Kalman filter.
    [pdf], [ps.gz]
  • Using EM to Estimate a Probability Distribution with a Mixture of Gaussians.
    [pdf], [ps.gz]
  • Deriving a marginal t-distribution.
    [pdf], [ps.gz]

Links: