Unsupervised Data Selection for Supervised Learning

TL;DR

This paper explores the potential of unsupervised data selection to improve supervised learning models, hypothesizing that better data quality can enhance generalization, though initial results are inconclusive.

Contribution

It introduces the idea of unsupervised data selection for supervised learning, proposing a methodological approach to improve data quality and model generalization.

Findings

Preliminary results are not robust.

Unsupervised data selection may enhance model generalization.

Further research is needed to validate the approach.

Abstract

Recent research put a big effort in the development of deep learning architectures and optimizers obtaining impressive results in areas ranging from vision to language processing. However little attention has been addressed to the need of a methodological process of data collection. In this work we hypothesize that high quality data for supervised learning can be selected in an unsupervised manner and that by doing so one can obtain models capable to generalize better than in the case of random training set construction. However, preliminary results are not robust and further studies on the subject should be carried out.