Smooth-pursuit learning in a closed-loop cortex - basal-ganglia model with spiking neurons
CBN (Computational Biology and Neurocomputing) seminars
Friday 11 April 2014
to 11:00 at
Bernhard Kaplan (CB/CSC/KTH)
Pursuit eye movements require an active feedback loop between stimulus perception and the selection of the appropriate eye motor command to track a moving stimulus. The question how the action selection takes place and more specifically how the circuitry involved in pursuit eye movements develops is not resolved.
In this seminar, a multi-layer spiking neural network model for the development of pursuit eye movements is presented. Our main hypothesis is that the Basal Ganglia (BG) are crucial for the development of eye movements and that this learning process can be described as a form of habit formation.
The model operates in a closed-loop between sensory perception and action-selection and is based on a model of the BG which enables the system to learn and adapt to changes in the environment. A sensory layer inspired by retinotopic cortical areas forwards information about stimulus speed, direction and position to the BG which in turn selects a motor command initiating an eye movement which dynamically changes the perceived retinal stimulus. I will present results using a supervised learning paradigm and discuss future directions including unsupervised development of the pursuit pathway and the role of delays in this closed-loop system.