[Illinois] MCB 493 Neural Systems Modeling
Category
Published on
Abstract
The purpose of this independent study is to give students hands-on experience in using computers to model neural systems. A neural system is a system of interconnected neural elements, or units. Students will use existing computer programs which will simulate real neural systems. They will compare the behavior of the model units with neurophysiological data on real neurons. The neural system models will all perform a useful computation, and the similarity between the behaviors of model units and real neurons will give students insight into how the real nervous system may actually work.
Bio
Professor Thomas Anastasio is an associate professor in the University of Illinois Department of Molecular and Integrative Physioloy and a full-time faculty member of the Beckman Institute. His main research interest is in Systems Neurobiology, which he defines as the application of conceptual and computational methods to complex and multilevel problems in neurobiology. He is also involved in teaching, writing, and technology transfer.
My overarching professional interest is to understand the interactions that sustain brain function as they occur on the molecular, neural, and social levels. Until recently my research has mainly concerned computational modeling at the neural systems level, and I have written a textbook on this topic. My current research focuses on computer modeling of the molecular interactions that underlie brain function in health and disease.
Professor Anastasio's recent scholarly work involves exploring analogies between neural and social systems. He has taught in all areas of neuroscience but his current teaching is divided between an advanced course on neural systems modeling and a survey course on society and the brain. He maintains an interest in technology transfer, which began with the development of a self-aiming camera that was designed on the basis of a neural system model. He looks forward to future work for new insights derived from further multilevel explorations.
Credits
Filming: George Daley
Editing: George Daley
Cite this work
Researchers should cite this work as follows:
Location
University of Illinois at Urbana-Champaign, Urbana, IL
Submitter
NanoBio Node, Obaid Sarvana, George Daley
University of Illinois at Urbana-Champaign
Tags
Lecture Number/Topic | Online Lecture | Video | Lecture Notes | Supplemental Material | Suggested Exercises |
---|---|---|---|---|---|
[Illinois] MCB 493 Lecture 1: Vectors, Matrices, and Basic Neural Computations | View HTML |
View | Simulation Tool |
||
Using mathematical and computational methods to simulate many aspects of neural systems function. |
|||||
[Illinois] MCB 493 Lecture 2: Recurrent Connections and Simple Neural Circuits | View HTML |
View | Simulation Tool |
||
Networks with recurrent connections, forming circuits, and containing only a few neural units can shape signals in time, produce oscillations, and simulate certain forms of low-level motor control. |
|||||
[Illinois] MCB 493 Lecture 3: Forward and Recurrent Lateral Inhibition | View HTML |
View | Simulation Tool |
||
Networks with forward and recurrent laterally inhibitory connectivity profiles can shape signals in space and time, and simulate certain forms of sensory and motor processing. |
|||||
[Illinois] MCB 493 Lecture 4: Covariation Learning and Auto-Associative Memory | View HTML |
View | Simulation Tool |
||
Networks with recurrent connection weights that reflect the covariation between pattern elements can dynamically recall patterns and simulate certain forms of memory. |
|||||
[Illinois] MCB 493 Lecture 5: Unsupervised Learning and Distributed Representations | View HTML |
View | Simulation Tool |
||
Unsupervised learning algorithms, given only a set of input patterns, can train neural networks to form distributed representations of those patterns that resemble brain maps. |
|||||
[Illinois] MCB 493 Lecture 6: Supervised Learning and Non-Uniform Representations | View HTML |
View | Simulation Tool |
||
Supervised learning algorithms can train neural networks to associate patterns and simulate the non-uniform distributed representations found in many brain regions. |
|||||
[Illinois] MCB 493 Lecture 7: Reinforcement Learning and Associative Conditioning | View HTML |
View | Simulation Tool |
||
Reinforcement learning algorithms can simulate certain types of associative conditioning and train neural networks to form non-uniform distributed representations. |
|||||
[Illinois] MCB 493 Lecture 8: Information Transmission and Unsupervised Learning | View HTML |
View | Simulation Tool |
||
Unsupervised learning algorithms can train neural networks to increase the amount of information they contain about their inputs and simulate the properties of sensory neurons. |
|||||
[Illinois] MCB 493 Lecture 9: Probability Estimation and Supervised Learning | View HTML |
View | Simulation Tool |
||
Supervised learning algorithms can train neural units and networks to estimate probabilities and simulate the responses of neurons to multisensory stimulation. |
|||||
[Illinois] MCB 493 Lecture 10: Time Series Learning and Nonlinear Signal Processing | View HTML |
View | Simulation Tool |
||
[Illinois] MCB 493 Lecture 11: Temporal-Difference Learning and Reward Prediction | View HTML |
View | Simulation Tool |
||
Temporal-difference learning can train neural networks to estimate the future value of a current state and simulate the responses of neurons involved in reward processing. |
|||||
[Illinois] MCB 493 Lecture 13: Predictor-Corrector Models and Probabilistic Inference | View HTML |
View | Simulation Tool |
||
[Illinois] MCB 493 Lecture 14: Future Directions in Neural Systems Modeling | View HTML |
View | |||
In the future, neural systems models will become increasingly complex and will span levels from molecular interactions within neurons to interactions between networks |