PROJECTS AT PSPL FINISHED IN 2006


Biologically Motivated Neural Network Architectures

by Gala Tsimerman
Supervised by Karina Odinaev & Igal Raichelgauz


Abstract

In the project, we tried to achieve collaboration between the electrical engineering faculty and the faculty of biology, by combining methods of electrical engineering with the knowledge of biology. The goal was to combine engineering techniques with biological studies. The assumption was that more realistic (biological) network would give better results in the electrical techniques (smaller errors).


The background

The neocortex constitutes nearly 80% of the human brain and is made of a repeating stereotypical microcircuit of neurons. The same microcircuit of neurons can simultaneously partake in an unrestricted number of tasks. This capability allows the neocortex to be parcelled into multiple overlapping functional vertical columns that form a foundation of functional compartmentalization of neocortex. The computational units of cortical circuits connected in the connectivity patterns.This is what we tried to learn about.

It appears that the combination of the engineering techniques with the biological relevance of studies holds many fruitful experimental avenues.


The final architectures that resolves from all the research:

1592 neurons.
6 layers. Each column is going through all the six layers.
Layer III receives input from other cortical columns.
Cells of layers II and III project to other parts of the cortex.
The scheme of the final architecture is represented below.



Tools

csim simulator, which is working on voice identification.


Conclusions

Based on the results, there is a great possibility that creation of biological network in the simulator will, in fact, improve the results.
Research in this field must go on.

Acknowledgment

I am grateful to my project supervisor Karina Odinaev for her guidance throughout this project and for her patience to work with a student that came from the faculty of Biology, with no previous knowledge in the engineering field.
I also thank Johanan Erez for all his help.

In addition I am grateful to the Ollendorf Minerva Center Fund for supporting this project.


RELATED DOCUMENTATION