Kalyan Veeramachaneni, Owen Derby, Dylan Sherry, and Una-May O’Reilly examine the challenge of producing ensembles of regression models for large datasets. The authors present a regression ensemble system, which produces model ensembles by relying on a GP-based learner. Their system pairs the stochastic nature of GP with a probabilistic parameter factoring method and the variable computation speed of the cloud to produce many diverse models for fusion. They discuss multiple robust strategies for fusing predictions from these ensembles of models. The paper is entitled “Learning Regression Ensembles with GP at Scale“.
Ensemble-based methods have several advantages. Related papers can also be downloaded from SPOTSeven’s web page, e.g., Comparing Ensemble-based Forecasting Methods for Smart-Metering Data, Ensemble-Based Modeling, Ensemble Based Optimization and Tuning Algorithms.