Neurox is a platform for processing neural networks. It handles common subsystems such as real time simulation, network state loading and storing from a living central repository, neuron state information. C++ thin client provides active processing of neural network, while a SQL backend does the living processing of the network performing slower network mobility calculations and state retention. Real-time simulation is imperative for a neural network to sustain, with out it cyclical systems will never form and the network will spiral itself into a dead calm. Implemented using a two state machine in a cellular automata environment, processing is based on first state, with results written to second state. When sequential processing is finished first and second states are swapped, thus incrementing virtual time. Real time is achieved by taking the derivative of time. The living network is persisted in a SQL server where it can be loaded from and saved to. While in database slower heavy calculations are made influencing mobility of neurons and persistence of heavily used network links. Dendrites can reconnect or seek new neurons, neurons can translate space, and other spatial transactions can be preformed. These changes will be translated back to the processing units on the client. Current processing state is also stored for persistence between application uses. Neurons are the main processing elements in this network and are custom programmed using an API. Implemented as dlls, they are loaded in to the network. Multiple concurrent neuron types can coexists in the network and use chemical interchanges at the synapses to communicate between these incongruent systems.

September 22, 2006
note: small posting, will work flushing out later.

Living Processing

Neurons have an axons and dentrites. Link to neuron information Neron - Box which has ins and outs, dentrites and axon/buttons respectivly. a box's outs are connected to another's in. A collection of nerons is a Grid, and a collection of grids is called Space.