Modeling the retinal horizontal cell layer on a massively parallel processor: a detailed neural network model

The authors describe the use of a massively parallel processor, the Connection Machine model CM-2, to simulate light responses of the horizontal cell network of the vertebrate outer retina. The network model is biophysically detailed; properties of all nonlinear voltage-gated membrane currents and intracellular calcium buffering mechanisms are modeled. Implementation and efficiency of execution on CM-2 and Cray supercomputers is discussed. Computational properties of the horizontal cell network under light and dark adapted conditions are analyzed. Results demonstrate that an increase in cell coupling following light adaptation improves the temporal resolution of the network at the expense of decreased spatial resolution. These changes may perform matched filtering, adjusting the response kinetics of the horizontal cell network to match those of cone photoreceptors over a range of background light levels