Local Synaptic Rules with Maximal Information Storage Capacity

A mathematical model of a natural process is always a kind of caricature. Today the physicists’ most popular model for the process of neural activation in the brain is the so-called spin-glass model — a mathematical model that is derived from theoretical investigations on Ising spins and has a certain formal similarity to mathematical models of brain dynamics. This type of model is certainly a long way from the reality of magnetic materials, but it is still further remote from real brain dynamics. On the other hand, it has often proved to be beneficial for a theoretical insight to step back from too close a look at reality and to work with caricatures. It is in this spirit that we want to compare and analyse a family of mathematical neural network models that investigate learning and synaptic plasticity (e.g. [1],[3],[4]).

[1]  H. C. LONGUET-HIGGINS,et al.  Non-Holographic Associative Memory , 1969, Nature.

[2]  G Palm,et al.  Computing with neural networks. , 1987, Science.

[3]  J J Hopfield,et al.  Neural networks and physical systems with emergent collective computational abilities. , 1982, Proceedings of the National Academy of Sciences of the United States of America.