Neuroscience 2005 Abstract
| Presentation Number: | 590.4 |
|---|---|
| Abstract Title: | The role of recurrent connections in the processing of reach movements in the parietal cortex. |
| Authors: |
Brozovic, M.*1
; Gail, A.1
; Andersen, R. A.1
1California Inst. of Technology, Pasadena, CA |
| Primary Theme and Topics |
Sensory and Motor Systems - Visuomotor Processing -- Spatial memory and sensorimotor transformations |
| Secondary Theme and Topics | Cognition and Behavior<br />- Animal Cognition and Behavior<br />-- Associative, nonassociative, and skill learning |
| Session: |
590. Eye and Reaching Movements: Cortical Processing Slide |
| Presentation Time: | Tuesday, November 15, 2005 8:45 AM-9:00 AM |
| Location: | Washington Convention Center - Room 201 |
| Keywords: | parietal cortex, neural nets, recurrent connections |
We present a theoretical study on the role of recurrent connections in processing sensorimotor transformations in the posterior parietal cortex. An experimental study (A. Gail SfN 2005) shows that activity in the monkey parietal reach region (PRR) represents the spatial goal of reach movements rather than the location of the visual cue during an anti-reach task. A 2-dimensional, 3-layer neural network (Zipser-Andersen type) with 2 different kinds of recurrent connections was trained to represent reach goals in its output layer, depending on the location of a visual cue and the behavioral context (reach/anti-reach) fed into its input layer. The network was trained using a backpropagation-through-time algorithm allowing the network to simulate memory-guided reach planning.
In the first version of the model only the units in the hidden layer were recurrently connected. The units in the hidden layer (‘PRR’) developed receptive fields with a spatial tuning representing the visual cue that was gain modulated with respect to the task being either a reach or an anti-reach. This means, although the network produced temporal dynamics and the desired output mapping, it could not explain experimental finding of motor-like tuning in PRR.
In a second model version we introduced additional recurrent feedback connections between the output and the hidden layer units. Again the network learned the proper input-output transformations. But this time the activity of the units in the hidden layer additionally reproduced the experimentally observed dependence of the spatial tuning from the movement goal instead of the visual cue. In conclusion, we suggest that the encoding of movement goals in the parietal cortex depends on top-down projections of motor-related activity from frontal areas during the learning of contex-dependent, spatial visuomotor associations.
In the first version of the model only the units in the hidden layer were recurrently connected. The units in the hidden layer (‘PRR’) developed receptive fields with a spatial tuning representing the visual cue that was gain modulated with respect to the task being either a reach or an anti-reach. This means, although the network produced temporal dynamics and the desired output mapping, it could not explain experimental finding of motor-like tuning in PRR.
In a second model version we introduced additional recurrent feedback connections between the output and the hidden layer units. Again the network learned the proper input-output transformations. But this time the activity of the units in the hidden layer additionally reproduced the experimentally observed dependence of the spatial tuning from the movement goal instead of the visual cue. In conclusion, we suggest that the encoding of movement goals in the parietal cortex depends on top-down projections of motor-related activity from frontal areas during the learning of contex-dependent, spatial visuomotor associations.
Supported by NIH, ONR
Sample Citation:
[Authors]. [Abstract Title]. Program No. XXX.XX. 2005 Neuroscience Meeting Planner. Washington, DC: Society for Neuroscience, 2005. Online.
Copyright © 2005-2026 Society for Neuroscience; all rights reserved. Permission to republish any abstract or part of any abstract in any form must be obtained in writing by SfN office prior to publication.
