Computational Maps in the Visual Cortex
     Figure 11.5
MiikkulainenBednarChoeSirosh
Home    
About the Authors
Back Cover    
Table of Contents 
Sample Chapter 
Figures    
References    
Errata    
Demos     
Talks/Courses 
Software    
Credits    
Purchase online at:

springeronline.com
amazon.com

Click on the image to see a PDF version (for zooming in)

Fig. 11.5. Long-range lateral connections in GMAP. The long-range excitatory lateral connection patterns for four sample neurons in GMAP are shown on top, located in iso-orientation patches as shown in the map below. Similar plotting conventions are used as in Figure 5.12: The small black square identifies the neuron itself in both plots, and the white outline on the map indicates the extent of the lateral connections after self-organization and pruning; before self-organization the lateral connections covered the whole map, as shown by the black square outline on top. The color coding in the top plots represents the target neuron's orientation preference, selectivity, and connection strength, and the map below encodes orientation and selectivity. The histogram in the middle shows the distribution of the target neurons' orientation preferences. Each neuron is most strongly connected to its closest neighbors; the long-range connections are patchy and connect neurons with similar orientation preferences. They extend longer than those in Figure 5.12 because more elongated input patterns were used during self-organization. As in LISSOM, these connections extend along the orientation preference of the source neuron: (a) 2o red, (b) 51o purple, (c) 91o light blue, and (d) 136o light green. They are narrow around the neuron but wider farther away. As will be seen in Chapter 13, specific connection patterns like these are crucial for perceptual grouping such as contour integration.