Efficient learning of nonlinear prediction models with time-series privileged information

TL;DR

This paper introduces methods for efficient nonlinear prediction in time-series models using privileged information, extending theoretical guarantees and demonstrating empirical benefits in limited data scenarios.

Contribution

It generalizes previous linear results to nonlinear latent systems and proposes algorithms for cases with unknown observation mappings.

Findings

Privileged time-series data improves nonlinear prediction accuracy.

Theoretical guarantees hold for known and unknown observation mappings.

Empirical results confirm the advantage of using privileged information.

Abstract

In domains where sample sizes are limited, efficient learning algorithms are critical. Learning using privileged information (LuPI) offers increased sample efficiency by allowing prediction models access to auxiliary information at training time which is unavailable when the models are used. In recent work, it was shown that for prediction in linear-Gaussian dynamical systems, a LuPI learner with access to intermediate time series data is never worse and often better in expectation than any unbiased classical learner. We provide new insights into this analysis and generalize it to nonlinear prediction tasks in latent dynamical systems, extending theoretical guarantees to the case where the map connecting latent variables and observations is known up to a linear transform. In addition, we propose algorithms based on random features and representation learning for the case when this map is unknown. A suite of empirical results confirm theoretical findings and show the potential of using privileged time-series information in nonlinear prediction.