Please, go to the NeuroOverview.pdf
file in this repository for an overview of this project.
Description of an ongoing personal project to develop a flexible system integrating different approaches to solve a generic problem, having the nervous system as an inspiration. The system is iterative, dynamically partitioning the input space-time, and assigning each partition to a specific module (Mixture of Experts), picked on the basis of its performance. This allows to combine multiple strategies to solve the same problem, providing improved performance, fault tolerance, efficiency, and reducing its overall cost. The modules are completely generic, e.g. from hard coded sets of rules to time-dependent neural networks (e.g. Spiking Neuron Model). Genetic algorithms are used for configuration, optimization, network structure design, formula evolution; focus on efficiency; extraction of symbolic information from raw data.
A more detailed presentation is available upon request.