Connected cortical recurrent networks
A model of an associative memory composed of many modules working as attractor neural networks with features of biological realism is proposed and analyzed using standard statistical physics techniques. The memories of the system are stored in the synapses between neurons in the same module and the synapses between neurons in different modules provide, the associations between these memories. A study of the memory storage properties as a function of the strength of the associations is performed and it is found that, if it is large, global retrieval phases can be found in which selective sustained actibities induced in modules which have not been stimulated. The form of the associations is such that, in the case of a tri-modular network studied, results from a psychophysical experiment on the simultaneous processing of contradictory information  can be qualitatively reproduced, within the limitations imposed by the simplicity of the model.
Unable to display preview. Download preview PDF.
- 9.Renart, A., Parga, N., Rolls, E. T.: Backprojections in the cerebral cortex: implications for memory storage. Neural Computation (to appear)Google Scholar
- 10.Renart, A., Parga, N., Rolls, E. T.: Associative memory properties of multiple cortical modules. Network: Computation in Neural Systems (submitted)Google Scholar