REBAR: Retrieval-Based Reconstruction for Time-series Contrastive Learning
This addresses the challenge of identifying meaningful positive pairs in time-series data, where invariances are less obvious than in vision, offering a novel solution for domain-specific applications.
The paper tackles the problem of constructing positive pairs for self-supervised contrastive learning in time-series data by proposing REBAR, a retrieval-based reconstruction measure, which achieves state-of-the-art performance on downstream tasks across various modalities.
The success of self-supervised contrastive learning hinges on identifying positive data pairs, such that when they are pushed together in embedding space, the space encodes useful information for subsequent downstream tasks. Constructing positive pairs is non-trivial as the pairing must be similar enough to reflect a shared semantic meaning, but different enough to capture within-class variation. Classical approaches in vision use augmentations to exploit well-established invariances to construct positive pairs, but invariances in the time-series domain are much less obvious. In our work, we propose a novel method of using a learned measure for identifying positive pairs. Our Retrieval-Based Reconstruction (REBAR) measure measures the similarity between two sequences as the reconstruction error that results from reconstructing one sequence with retrieved information from the other. Then, if the two sequences have high REBAR similarity, we label them as a positive pair. Through validation experiments, we show that the REBAR error is a predictor of mutual class membership. Once integrated into a contrastive learning framework, our REBAR method learns an embedding that achieves state-of-the-art performance on downstream tasks across various modalities.