To develop an appropriate internal representation, a deterministic learning algorithm that can adjust not only weights but also the number of adopted hidden nodes is proposed. The key mechanisms are 1. (1) the recruiting mechanism that recruits proper extra hidden nodes, and 2.(2) the reasoning mechanism that prunes potentially irrelevant hidden nodes. This learning algorithm can make use of external environmental clues to develop an internal representation appropriate for the required mapping. The encoding problem and the parity problem are used to demonstrate the performance of the proposed algorithm. The experimental results are clearly positive.
Relation:
Mathematical and Computer Modelling, 38(3), 339-350
