New theory for activity in hierarchical neural networks
A p-adic statistical field theory describes how activity spreads through tree-like structures.
The Buice-Cowan field theory [PRE 75, 051919 (2007)] describes fluctuations and correlations of neural activity. This theory uses a path integral where the mean-field limit is the Wilson-Cowan equation.
In the continuum limit, neurons are points in a Euclidean space. Activity spreads by ordinary diffusion in this space, which does not encode hierarchical organization or small-world properties.
A new version of the theory uses an infinite, self-similar, tree-like structure described by p-adic numbers. This hierarchical network is small-world when the decay exponent is smaller than the dimension.
The linear response in the p-adic theory uses a nonlocal p-adic diffusion operator. For couplings that decay by a fixed factor per level, this is the Taibleson-Vladimirov operator.
The critical theory is an ultrametric Reggeon field theory. Its upper critical dimension is twice the decay exponent of the coupling.
Simulations on a tree show the avalanche-size exponent is when coupling decays slowly. The exponent decreases toward the directed-percolation value when coupling decays fast.
Key facts
- The p-adic theory uses an infinite, self-similar, tree-like structure.
- The decay of the coupling up the hierarchy decides whether mean-field exponents apply.
- The hierarchical network is small-world if the decay exponent is smaller than the dimension.