Year: 2014
Type: article
Abstract: The brain encodes the space in various reference frames.The key role in spatial transformations is played by the posterior parietal cortex where neurons combine retinal location of visual stimulus with gaze direction to encode spatial information.This nonlinear dependence of neuronal responses, gain modulation, is considered a fundamental computational principle used in the brain.The important insight can be obtained through computational models, typically artificial neural networks.In this paper, we test the Zipser-Andersen model but use more realistic and variable stimuli, employing the simulated iCub robot.The multi-layer perceptron was able to successfully perform coordinate transformation from eye-to body-centered reference frame, using gaze information.Model achieves high accuracy of 2 to 4 degrees on testing data, depending on the dataset variability.We provide visualisation techniques for analysing the network, and the effects of gain modulation and shifting receptive fields.Our results confirm previous findings that hidden neurons use various intermediate codings that mediate transformations.
Authors: Marcel Svec, Igor Farkaš
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2401616697
