Alireza Goudarz

Complex Systems Summer School 2014 Home Surveys/Evaluations Logistics Payment Instructions Faculty Lecture Slides Panel Questions Panel Responses Participants 1-Minute Summaries Schedule Background Readings Software Projects & Working Groups Project Group Meeting Schedule Sign Up Sheet for Meetings with Miguel and Josh Tutorials Blog Project Presentations Final Papers After Hours Restaurant Guide Resource Guide Photos TShirts

Email: alireza.goudarzi at gmail.com

url: Homepage

I am a second year PhD computer science student in the University of New Mexico. My broad research area is computation and learning in dynamical systems with application to computation using biomolecular systems and self-assembled nanoscale networks. Align with my broad research area, I am interested in machine learning, data mining, dynamical system theory, artificial intelligence, computational learning theory, physics of disordered systems, complex networks and related fields.

Currently, my focus is on reservoir computing, which is a new paradigm in neural networks and has shown great performance in non-linear and chaotic signal processing. In contrast to traditional neural networks, in which the system is fine tuned using a learning algorithm to solve a particular task, in reservoir computing a recurrent neural network is initialized and fixed with no further fine tuning. The signal is given to the network and a linear readout layer performs a linear combination of the states to produce the target output. The weights of the linear combination is calculated using linear regression which can be performed in closed form efficiently. My goal is to understand this paradigm better and study what kinds of problems they can and they cannot solve and how the paradigm can be applied to unconventional computing substrates such as nanoscale self-assembled electronics or DNA molecules.

My most recent work on reservoir computing is the direct estimation of error in task solving based on the structure of the dynamical system and statics of the input and output. The approach is currently limited to linear systems but I'm hoping to extend it to nonlinear systems and systems with finite dynamic range. I would be happy to talk about this more to other summer school students and even collaborate on taking it further.

Selected publications:

• [Emergent Criticality through Adaptive Information Processing in Boolean Networks]
• [Computational capabilities of random automata networks for reservoir computing]
• [DNA Reservoir Computing: A Novel Molecular Computing Approach]
• [Reservoir Computing Approach to Robust Computation using Unreliable Nanoscale Networks]

---

Skills to contribute to the summer school teams:

• Programming in C/C++, Python, Matlab, Octave, netlogo
• Familiar with ODE, Linear algebra, information theory, stochastic processes, machine learning, neural network, complex networks, dynamical systems, evolutionary algorithms, simulated annealing, and agent-based modeling.