Modeling Time Series Similarity with Siamese Recurrent Networks
This work addresses the challenge of time series similarity for applications like biometric recognition, offering a supervised learning approach that improves over traditional handcrafted measures and unsupervised methods, though it is incremental as it combines existing ideas from time-series modeling and metric learning.
The paper tackles the problem of measuring similarities between time series by proposing siamese recurrent networks (SRNs) that learn a similarity measure through a classification loss, demonstrating their effectiveness in tasks like signature, voice, and sign language recognition with few examples per class.
Traditional techniques for measuring similarities between time series are based on handcrafted similarity measures, whereas more recent learning-based approaches cannot exploit external supervision. We combine ideas from time-series modeling and metric learning, and study siamese recurrent networks (SRNs) that minimize a classification loss to learn a good similarity measure between time series. Specifically, our approach learns a vectorial representation for each time series in such a way that similar time series are modeled by similar representations, and dissimilar time series by dissimilar representations. Because it is a similarity prediction models, SRNs are particularly well-suited to challenging scenarios such as signature recognition, in which each person is a separate class and very few examples per class are available. We demonstrate the potential merits of SRNs in within-domain and out-of-domain classification experiments and in one-shot learning experiments on tasks such as signature, voice, and sign language recognition.