Co-adaptive learning over a countable space

TL;DR

This paper analyzes co-adaptive learning in an online, closed-loop setting, providing theoretical guarantees that it can outperform fixed decoders under certain conditions, with applications in BCI and other fields.

Contribution

It offers the first general theoretical analysis of co-adaptive learning in a countable space, establishing performance guarantees in an online setting.

Findings

Co-adaptive learning can outperform fixed decoders with high probability.

Theoretical guarantees depend on a specific condition being satisfied.

Applications include brain-computer interfaces and adaptive systems.

Abstract

Co-adaptation is a special form of on-line learning where an algorithm $\mathcal{A}$ must assist an unknown algorithm $\mathcal{B}$ to perform some task. This is a general framework and has applications in recommendation systems, search, education, and much more. Today, the most common use of co-adaptive algorithms is in brain-computer interfacing (BCI), where algorithms help patients gain and maintain control over prosthetic devices. While previous studies have shown strong empirical results Kowalski et al. (2013); Orsborn et al. (2014) or have been studied in specific examples Merel et al. (2013, 2015), there is no general analysis of the co-adaptive learning problem. Here we will study the co-adaptive learning problem in the online, closed-loop setting. We will prove that, with high probability, co-adaptive learning is guaranteed to outperform learning with a fixed decoder as long as a particular condition is met.