New methods of machine learning for the construction of integrated neuromorphic and associative-memory knowledge bases

There is now a consensus that information systems designed to solve problems in complex, dynamic domains will require intelligent use of sophisticated knowledge bases. The construction of such bases through explicit learning is difficult. As a result, various methods of machine learning from examples are tried to alleviate the problem. Experimental evidence is presented on the successful performance of two new learning methods in the acquisition of inverse dynamics models of robot manipulators.<<ETX>>