Time-line hidden Markov experts for time series prediction

摘要

A modularised connectionist model, based on the mixture of experts (ME) algorithm for time series prediction, is introduced. A group of connectionist modules learn to be local experts over some commonly appeared states in a time series. The dynamics for combining the experts is a hidden Markov process, in which the states of a time series are regarded as states of a HMM and each of them associates to an expert. However, the state transition property is time-variant and conditional on the dynamic situation of the time series.