Related Experiment Videos
A neural network study of precollicular saccadic averaging
K P Krommenhoek1, W A Wiegerinck
1Department of Medical Physics and Biophysics, University of Nijmegen, The Netherlands. karin@mbfys.kun.nl
Biological Cybernetics
|August 26, 1998
Summary
Saccadic averaging, where eye movements land between two targets, may occur upstream of the superior colliculus. A new neural network model demonstrates this averaging as an emergent property, offering testable predictions for future research.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Ophthalmology
Background:
- Saccadic averaging is a phenomenon where simultaneous visual stimuli result in an eye movement endpoint between the stimuli.
- Neural correlates suggest saccadic averaging occurs in or before the superior colliculus.
Purpose of the Study:
- To propose and analyze a novel neural network model for saccadic feedback.
- To investigate the mechanism of weighted saccadic averaging within this model.
- To determine if saccadic averaging can occur upstream of the superior colliculus.
Main Methods:
- Development of a neural network model with output units mimicking the superior colliculus and hidden layers representing the posterior parietal cortex.
- Simulation of simultaneous retinal target presentation with varying intensities.
- Analysis of emergent averaging properties and the role of linear/non-linear activation functions.
Main Results:
- The neural network model exhibits saccadic averaging as an emergent property.
- Averaging results from the linear dependence of hidden unit input on retinal error.
- Model predictions align with physiological data when retinal error input is normalized.
Conclusions:
- Saccadic averaging is likely an emergent property of neural networks involved in eye movement control.
- The proposed model suggests saccadic averaging may occur upstream of the superior colliculus.
- The model offers testable predictions regarding electrical stimulation and eye position effects on saccadic averaging.