Improving XGBoost with Imagination Sampling

Eric Holloway

doi:10.33014/issn.2640-5652.2.1.holloway.1

Eric Holloway

DOI: https://doi.org/10.33014/issn.2640-5652.2.1.holloway.1

Keywords: imagination sampling, gradient boosting, xgboost, mechanical turk, human in the loop, human computation, multibox, random forest, computation, oracle

Abstract

Imagination Sampling is the usage of a person as an oracle for generating or improving machine learning models. Previous work demonstrated a general system for using Imagination Sampling for obtaining multibox models. Here, the possibility of importing such models as the starting point for further automatic enhancement is explored.