Controlling the chaotic neural network-a way of information integration

Concerns the use of neural nets for sensor fusion and integration. We stabilized the unstable equilibrium state of a chaotic network by a small external signal. The network has the first-order and second-order random and diluted connections, and its dynamics can be stable, periodic and chaotic for different values of parameters. The famous OGY method for controlling the chaos is applied to the evolution equation of the network. The controlling algorithm is efficient and convenient. When the algorithm is used to the chaotic network, the iterating sequence is stabilized at the equilibrium point after several iterations and it becomes chaotic again when the control is disabled, as shown by the numerical experiments. The control process can be regarded as the information integration between the network modules while the controlling signal being regarded as the information from other modules. A two-module architecture for information integration is proposed based on the controlling method.