Hierarchical Mixtures of Experts and the EM Algorithm

TLDR

The architecture is based on a hierarchical mixture model in which both mixture coefficients and components are generalized linear models. The authors introduce a tree‑structured architecture for supervised learning. Learning is formulated as a maximum‑likelihood problem, solved with an EM algorithm and an online incremental variant. Simulations in the robot dynamics domain demonstrate the method’s performance.

Abstract

We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain.