Computational Neuroscience MBB Track
Contact Information
Website: Computer Science
Advising
Prof. Stuart Shieberr, http://www.eecs.harvard.edu/~shieber
Ellen Holloway, ehollowa@seas.harvard.edu
The Intellectual Basis
Computational neuroscience attempts to understand mental abilities -- such as perception, language, motor control, and learning -- by designing artificial systems with similar capabilities. It uses tools from computer science, mathematics, and engineering to construct theories which can then be implemented by computer programs or by special purpose hardware. The current state of the art includes systems for face recognition, speech recognition, and language understanding, and autonomous land vehicles. The theories developed in computational neuroscience are often motivated by findings about biological systems and, in turn, they provide theoretical models for psychologists and biologists to investigate. Some computational theorists explicitly design their theories to take into account the known properties of neurons. One can, for example, model the brain as a complex neural network and investigate how many memories can be stored, how many patterns can be learned, and what types of neuronally plausible algorithms are most effective for these tasks. Or one can design a theory of vision based on the known properties of the visual cortex. By contrast, other theories might bypass the neuronal level and concentrate on describing the performance of the system. Such a theory might predict, for example, under what conditions face recognition is possible and what types of faces might be hardest for a human to recognize. Computational neuroscience is interdisciplinary and there is a growing, and very exciting, interaction between it and cognitive neuroscience.