Statistical Mechanics of Linear and Nonlinear Time-Domain Ensemble Learning

摘要: Conventional ensemble learning combines students in the space domain. In this paper, however, we combine time domain and call it time-domain learning. We analyze, compare, discuss generalization performances regarding of both a linear model nonlinear model. Analyzing framework online using statistical mechanical method, show qualitatively different behaviors between two models. model, dynamical error are monotonic. analytically that is twice as effective conventional Furthermore, features nonmonotonic when rate small. numerically performance can be improved remarkably by phenomenon divergence