|
  |
|
|
Krasnow Institute > Monday Seminars > Abstracts Qualitative Reasoning as a Modeling Tool for Computational Neuroscience Jeffrey L. Krichmar A novel approach to modeling neural behavior is presented using a "qualitative reasoning" algorithm. This paper introduces the Qualitative Reasoning Neuron(QRN), which is capable of qualitatively reproducing single neuron behavior, but is computationally simple enough to use in large scale neural networks without loss of critical details. QRN simulates the behavior of a neuron by qualitatively describing the relationships among the various ionic currents and cell's membrane potential. Since the interaction between parameters is "reasoned", no quantitative values need to be calculated. The lack of complex calculations makes the QRN algorithm extremely efficient. In the present study, QRN is used to model a single cerebellar Purkinje cell. Two QRN models are developed: The first matches the number and type of channels and the number and type of compartments to a recent GENESIS model of the Purkinje cell (De Schutter& Bower 1994). This model shows a two-fold improvement in computer processing time over the comparable GENESIS model. The second model has approximately the same number of compartments as the first, but simplifies extraneous channels. This second model shows a six-fold improvement in computer processing time over the GENESIS model. The output of the QRN models are testable, accurate and quantitative. The complex spike time course, simple spike firing rates (overall mean [+or- S.D.] 36.2 +or- 16.1 Hz for the first model and 59.0 +or- 20.2 Hz for the second model), interspike intervals (overall mean [+or- S.D.] 26.7 +or- 21.2 ms for the first model and 16.3 +or- 5.9 ms for the second model) are all within physiological ranges.
The Krasnow Institute for Advanced Study |
|
