Amir massoud Farahmand

Ph.D. Student
Department of Computing Science
University of Alberta
2-21 Athabasca Hall
Edmonton, Alberta
Canada T6G 2E8
email: Reverse rima and add @cs, then a dot and then ualberta and then dot ca!

Reverse rima and add @cs, then a dot and then ualberta and then dot ca!

News!

Our NIPS paper on Regularized Policy Iteration has been accepted. I'll provide more information soon! (Oct 2008)
Towards Manifold-Adaptive Learning has been presented in the Topology Learning workshop at NIPS. The main message is that many conventional machine learning methods may actually be manifold-adaptive, i.e. if seemingly high-dimensional data actually come from a low-dimensional manifold, such a manifold-adaptive method behaves as if it is dealing with a low-dimensional problem. We presented two manifold-adaptivity results: one for dimension estimation and another for the traditional K-NN regressor. (Dec 2007)

Our Global Visual-Motor Estimation for Uncalibrated Visual Servoing has been accepted and presented in the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2007. (Oct 2007)

Our Manifold-Adaptive Dimension Estimation paper has been accepted and presented in the International Conference on Machine Learning (ICML) 2007. In this paper, we provide an algorithm for estimating the intrinsic dimension of a manifold embedded in a possibly high-dimensional space. We prove a convergence rate for this method. The good news is that the convergence rate depends only on the intrinsic dimension of the manifold and not the dimension of the embedding space. Solving the curse of dimensionality?! Not yet, but a step along that way! (; (July 2007)

Research Interests

Regularities in Reinforcement Learning (2007-present)

Nonparametric L2 Regularized Methods for Action-Value Estimation

Manifold Learning (2006-present)
Application of Machine Learning Methods in Visual-Servoing Problems (2006-present)
Incremental Basis Generation in Reinforcement Learning (2006)
Hierarchical [Reinforcement] Learning in Behavior-based Systems (MS. Thesis) (2003-2005)

Structure Learning
Behavior Learning
Concurrent Behavior and Structure Learning
Behavior Evolution
Concurrent Behavior Evolution and Structure Learning
Automatic Generation of Reinforcement Signal

Evolutionary Robotics

Hybrid Evolution/Learning in Hierarchical Behavior-based Systems (2004-2005)
Evolving Neural Network-based Controllers for Robots (2001)

Investigating the Effect of Reward Uncertainty on the Value and the Policy of the Agent. (2004-2005)
Automatic Abstraction of Knowledge from Stimuli-Response information (2004-2005)

Structural Abstraction
Temporal Abstraction

Synchronization of Chaotic Systems (2003-4)
Stable Fuzzy Controllers (2003)
Combinatorial Optimization Problem (Channel Assignment Problem) using Chaotically-Enhanced Hopfield Neural Network (2002-3)
Calculating Resonant Frequencies of a Metallic Cavity using Finite Element Method (BS Thesis) (2001-2)
Investigating the Effect of Metallic Electrode and Buffer Layer on Dielectric Waveguides (Internship) (2001)
Several Researches on Evolutionary Computation (2000-current)

Evolutionary Robotics
Evolution of Hidden Markov Models (2001)
Distributed Evolutionary Systems (2000)
Electromagnetic Filter Design (2000)
Analog Circuits Design (2000)

My Machine Learning-related Surfings