Study of correlations of in a neural network
Neural networks function with a lot of randomness, nevertheless they are
quite successful in performing some defined tasks. It is therefore expected
there would be correlations between parts of the network, especially between
the parts that define the tasks and those that produce the results.
The aim of this project is to anaylyse the correlations bewteen different
parts of such a network in operation. One of the earliest references
(which uses analytical tools due to the simplicity of the model) is
"Theory of correlations in stochastic neural networks",
I. Ginzburg and H. Sompolinsky,
Phys. Rev. E 50, 3171 (1994).
The work to be carried out in this project involves familiarization with simple neural network models, construction of a computer program to simulate a model, and extract statistics in order to analyze the correlations.
The student is expected to have reasonably good computational skills. (Improvement of computational skills should also be seen as a part of the aims of this project.)