Active training on the CMAC neural network

The CMAC neural network presents a rigid architecture for learning and generalizing simultaneously, a limitation stressed with sparse or non-dense training datasets, and hardly solved by the current training algorithms. This paper proposes a novel training algorithm that overcomes the mentioned tradeoff. The training mechanism is based on the minimization of the stiffness energy of the output, solution based on the active deformable model theory. These ideas lead to a cell-interaction-based internal update mechanism that preserves the potential CMAC learning capabilities and delivers a higher generalization degree than the one a-priori embedded in the CMAC architecture. The training mechanism is derived entirely from a rigorous theoretical study. This analysis is supported with comparative results on the inverse kinematics of a robotic arm, which prove the excellent performance of the proposed active training.