On Contrastive Representations of Stochastic Processes

TL;DR

This paper introduces CReSP, a contrastive learning framework for stochastic process representations that avoids exact reconstruction, effectively handling high-dimensional and noisy data across various applications.

Contribution

The paper presents a unifying contrastive learning framework for stochastic processes that improves robustness and transferability over traditional reconstruction-based methods.

Findings

Effective for periodic functions, 3D objects, and dynamical processes.

Tolerates noisy high-dimensional observations better than traditional methods.

Learned representations transfer well to downstream tasks.

Abstract

Learning representations of stochastic processes is an emerging problem in machine learning with applications from meta-learning to physical object models to time series. Typical methods rely on exact reconstruction of observations, but this approach breaks down as observations become high-dimensional or noise distributions become complex. To address this, we propose a unifying framework for learning contrastive representations of stochastic processes (CReSP) that does away with exact reconstruction. We dissect potential use cases for stochastic process representations, and propose methods that accommodate each. Empirically, we show that our methods are effective for learning representations of periodic functions, 3D objects and dynamical processes. Our methods tolerate noisy high-dimensional observations better than traditional approaches, and the learned representations transfer to a range of downstream tasks.