next up previous contents index
Next: 7.6.2 Parallel Computers? Up: Parallel Computing in Previous: Parallel Computing in

7.6.1 What Is Computational Neurobiology?

Neurobiology  is the study of the nervous system. Until recently, most neurobiology research centered around exposing different neural tissue preparations to a wide range of environmental stimuli and seeing how they responded. More recently, the growing field of computational neurobiology has involved constructing models of how we think the nervous system works [Segev:89a], [Wehmeier:89a], [Yamada:89a]. These models are then exposed to a wide range of experimental conditions and their responses compared to the real neural systems. Those models that are demonstrated to accurately and reliably mimic the behavior of real neural systems are then used to predict the neural system's response to new and untried experimental situations, and to make firm predictions about how the neural system should respond if our theory of neural functioning is correct. Simplified models are also used to determine which features of a real neural system are critical to its underlying behavior and function. In doing so, they also indicate which features of the real system have no effect on desired system performance.

Computer modelling of neural structures from the level of single cells to that of large networks has, until recently, been rather isolated from mainstream experimental neurophysiology [Koch:92a]. Largely, this has been due to limitations of computer power which have necessitated reducing the models to such a basic level that their biological relevance becomes questionable. More powerful computer platforms, such as the parallel computers which have been used at Caltech, allow the construction of simulations of sufficient detail for their results to be compared directly with experimental results. Furthermore, the inherent flexibility of the modelling approach allows the neurophysiologist to observe the effect of experimental manipulations which are presently difficult or impossible to carry out on a biological basis. In this way, neural modelling can make firm predictions that can be confirmed by later experimentation [Bhalla:93a].



next up previous contents index
Next: 7.6.2 Parallel Computers? Up: Parallel Computing in Previous: Parallel Computing in



Guy Robinson
Wed Mar 1 10:19:35 EST 1995