Degrees: PHD, 2001, Univ Cincinnati Cincinnati; MS, 1996, Univ Timisoara; BS, 1995, Univ Timisoara

Bio:

Associate Editor for the IEEE Transactions on Neural Networks journal (since December 2005)
Co-organized workshops on Autonomous Robotics for high-school students participating in the FIRST robotics competition (in January 2006 and Fall 2006)

Research areas:

Neural models for individual and group brainstorming
Brainstorming is the process by which ideas are generated, either individually or in group. A common belief is that group brainstorming is more productive than individuals brainstorming alone. But research has shown the contrary: groups generate fewer ideas than individuals. We aim at developing neural models that can be used as a framework to study individual and group brainstorming and the effect of various cognitive and social influences on brainstorming.

Model extraction from trained neural networks
Neural networks are a widely used modeling tool. A network learns to approximate a function from some sample data and it generalizes well to unseen samples. But neural networks are also hard to understand, because there is no easy way to extract an explicit model out of a trained network. We developed an adaptive method to extract explicit piecewise linear models from trained neural networks and used it to extract analog circuits models.

Publications:

1. S. Doboli, A.A. Minai (2006). Latent attractors: A general paradigm for context-dependent neural computation. In Ke Chen, Lipo Wang (Eds.) Trends in Neural Computation, Springer-Verlag, ISBN-10: 3-540-36121-9
2. S. Doboli, A.A. Minai and P.J. Best (2000). Latent attractors: A model for context-dependent place representations in the hippocampus. Neural Computation, 12:1009-1043.
3. S. Doboli and A.A. Minai (2003). Network capacity analysis for latent attractor computation. Network:Comput. Neural Syst., 14:273-302.
4. H. Zhang, S. Doboli, H. Tang, A. Doboli (2005). Compiled code simulation of analog and mixed-signal systems using piecewise linear modeling of nonlinear parameters. Integration the VLSI Journal.

Courses taught

Computer Architecture, Artificial Intelligence, Database Systems, Robotics, Sensor Networks

Internship courses

NSF funded program for internships in entrepreneurship education together with Stony Brook University, Suffolk Community College, and SUNY Farmingdale.

Independent studies

1. Development of a web database application for a used on-line bookstore.
2. Robot navigation using reinforcement learning and neural networks.
3. Neural models for sequence learning.
4. Web application using JSP and SQL Server.
5. Swarm intelligence

Honors thesis for undergrads

Richard Bateman. Training a neural network to learn how to play Othello using reinforcement learning. (defended May 2004). Results of this thesis were presented at two student conferences. A paper was published in the Proceedings of the Conference on Computer Games: Design, AI and Education (CGAIDE'2004).Grad work supervised

1. Data mining for airline security applications
2. A study of the dynamics of a multi-agent system with application to trading markets
3. Negative Databases
4. Learning frequent tracks in a wireless sensor network
5. Review of geographic routing methods in wireless sensor networks