Continual Learning for Human State Monitoring

TL;DR

This paper introduces two new continual learning benchmarks for human state monitoring with time series data, evaluating existing strategies and highlighting challenges in knowledge accumulation over fixed subjects.

Contribution

It presents novel CL benchmarks tailored for real-world human monitoring scenarios and assesses the performance of existing strategies on these benchmarks.

Findings

Forgetting is easily mitigated with simple finetuning in these benchmarks.

Existing CL strategies struggle to accumulate knowledge over fixed test subjects.

Benchmarks reflect real-world domain-incremental challenges.

Abstract

Continual Learning (CL) on time series data represents a promising but under-studied avenue for real-world applications. We propose two new CL benchmarks for Human State Monitoring. We carefully designed the benchmarks to mirror real-world environments in which new subjects are continuously added. We conducted an empirical evaluation to assess the ability of popular CL strategies to mitigate forgetting in our benchmarks. Our results show that, possibly due to the domain-incremental properties of our benchmarks, forgetting can be easily tackled even with a simple finetuning and that existing strategies struggle in accumulating knowledge over a fixed, held-out, test subject.