September 2005
Volume 5, Issue 8
Free
Vision Sciences Society Annual Meeting Abstract  |   September 2005
A self-organizing neural network model of the development of motion direction selectivity, orientation, and ocular dominance maps and receptive fields in V1 and MT
Author Affiliations
  • Alexander M. Harner
    Neuroscience Program, Boston University, Boston, MA
  • Takeo Watanabe
    Neuroscience Program, Boston University, Boston, MA
Journal of Vision September 2005, Vol.5, 900. doi:10.1167/5.8.900
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Alexander M. Harner, Takeo Watanabe; A self-organizing neural network model of the development of motion direction selectivity, orientation, and ocular dominance maps and receptive fields in V1 and MT. Journal of Vision 2005;5(8):900. doi: 10.1167/5.8.900.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: Numerous neural models have been proposed to explain how major characteristics of ocular dominance (OD) and orientation selectivity (OS) maps and receptive fields (RFs) may develop in V1. Yet, few examine the development of other features in V1, and hardly any examine the development of features in higher visual areas. In 1999 and 2000 (ARVO), we presented the first models of the simultaneous development of motion direction selective (DS) maps and RFs in V1 and MT/V5, respectively, given moving bar stimuli. Here, we extend this model to examine the joint development of DS, OS, and OD maps and RFs in V1 and MT given noise and natural moving stimuli to simulate pre- and postnatal conditions. Methods: To learn multiple maps in multiple areas, we developed a new type of high-dimensional self-organizing map (SOM) involving multiple layers, parallel ‘winners’, and biologically realistic mechanisms. To model DS, each spatial connection contains multiple time-delayed connections and our unique principal-component-based, spatiotemporal learning rule. Results: Given stationary stimuli, the model develops joint, ocularly-matched OS and OD maps and RFs with most of the major characteristics observed in V1 (Blasdel, 1992) under both post- and prenatal conditions. It uniquely shows the automatic formation of OS singularities near OD band centers and local orthogonality between OS and OD. Given moving stimuli, the model produces DS maps and RFs with many of the characteristics recently observed in MT (Diogo et al, 2003; Livingstone et al, 2001) and V1 (Weliky et al, 1996; Emerson, 1997), including opposing DS fractures, which are unique and common in DS maps. Conclusion: These results demonstrate that computational principles employed for learning OS and OD (e.g. competitive Hebbian learning with faster-than-linear inhibition) can be generalized to learn complex features in higher areas with the addition of multiple layers and spatiotemporal learning.

Harner, A. M. Watanabe, T. (2005). A self-organizing neural network model of the development of motion direction selectivity, orientation, and ocular dominance maps and receptive fields in V1 and MT [Abstract]. Journal of Vision, 5(8):900, 900a, http://journalofvision.org/5/8/900/, doi:10.1167/5.8.900. [CrossRef]
Footnotes
 NSF 0418182, NIH R01 EY015980-01, Human Frontier Foundation RGP18/2004, Audrey Harner
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×