B1565
Title: On high-dimensional asymptotic properties of model averaging estimators
Authors: Ryo Ando - The University of Tokyo (Japan) [presenting]
Fumiyasu Komaki - RIKEN CBS (Japan)
Abstract: When multiple models are considered in regression problems, the model averaging method can be used to weigh and integrate the models. The purpose is to examine how the goodness-of-prediction of the estimator depends on the dimensionality of explanatory variables when using a generalization of the model averaging method in a linear model. The case of high-dimensional explanatory variables is specifically considered, with multiple linear models deployed for subsets of these variables. Consequently, the optimal weights that yield the best predictions are derived. It is also observed that the double-descent phenomenon occurs in the model averaging estimator. Furthermore, theoretical results are obtained in the case of adapting methods such as the random forest to linear regression models. Finally, a practical verification is conducted through numerical experiments.
