Validity analysis of chaos generated with Neural-Network-Differential-Equation for robot to reduce fish's learning speed

We have been investigating a method to compose some intelligent robots. Continuous catching and releasing experiment for fish induces the fish to find some escaping strategies staying at corners of a pool. To make fish-catching robot more intelligent than fish's adapting and escaping abilities, we have proposed a chaos-generator comprising Neural-Network-Differential-Equation (NNDE) and an evolving mechanism to have the system generate chaotic trajectories as many as possible. We believe that the fish could not be adaptive enough to escape from chasing net with chaos motions since unpredictable chaotic motions of net may go beyond the fish's adapting abilities to the net motions. In this report we confirmed that four type of chaos are generated by NNDE and one of them is valid for decreasing the fish's learning velocities, which we judged that could be thought that the chaos increased the robot's intelligence relatively against fish's escaping intelligence.