Method to Find Key Shapelets from Ripples in Hippocampus of Rats

Abstract

Since ripples in hippocampal CA1 diversify with episodic experience to increase information entropy, we hypothesized that ripples contain local waveforms (shapelets) that episodic memories are encoded. Finding shapelets from ripples of rats that experienced an episode contributes to verify our hypothesis. In this letter, to find key shapelets from ripples, we proposed a method consisting of k-shape for clustering time-series waveforms, shapelet transform to classify time-series waveforms, and L₁ regularized logistic regression to select key factors for classification in shapelet transform. Among these, k-shape and shapelet transform should be improved because k-shape has a restriction that lengths of time-series waveforms must be the same, and similarity criteria of time-series waveforms in the both methods are inconsistent. To solve the problems, we improved functions defined as similarity between time-series waveforms. Our experimental results showed that the proposed method was able to find key shapelets from ripples. © 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.