Slowly evolving geometry in recurrent neural networks I: extreme dilution regime
arXiv:cond-mat/0403615 · doi:10.1088/0305-4470/37/31/002
Abstract
We study extremely diluted spin models of neural networks in which the connectivity evolves in time, although adiabatically slowly compared to the neurons, according to stochastic equations which on average aim to reduce frustration. The (fast) neurons and (slow) connectivity variables equilibrate separately, but at different temperatures. Our model is exactly solvable in equilibrium. We obtain phase diagrams upon making the condensed ansatz (i.e. recall of one pattern). These show that, as the connectivity temperature is lowered, the volume of the retrieval phase diverges and the fraction of mis-aligned spins is reduced. Still one always retains a region in the retrieval phase where recall states other than the one corresponding to the `condensed' pattern are locally stable, so the associative memory character of our model is preserved.
arXiv comment18 pages, 6 figures
References in corpus (7)
- Finite Connectivity Attractor Neural Networks
- On the Hopf algebraic origin of Wick normal-ordering
- Coupled dynamics of fast spins and slow exchange interactions in the XY spin-glass
- Cluster derivation of Parisi's RSB solution for disordered systems
- The XY Spin-Glass with Slow Dynamic Couplings
- Hierarchical Self-Programming in Recurrent Neural Networks
- Supervised Learning with Restricted Training Sets: a Generating Functional Analysis