ICE - an Incremental Hybrid System for Continuous Learning

ICE is a new incremental construction algorithm of a hybrid system for continuous learning tasks. The basis of the hybrid system is a radial basis function (RBF) network layer. The second layer consists of local models. The two layers are closely combined with a strong interaction. For example information from the model-layer is used by the RBF-layer to decide if new RBF-neurons are needed and the model layer uses the RBF-neurons as the base for its models. The number of RBF-neurons and the number of local models have not to be determined in advance, which is one of the main advantages of ICE. Another advantage over existing methods are useful network outputs already during the initial learning phase. The development of the described approach is embedded in a larger project that is primarily concerned with system identification tasks for industrial control such as steel processing.