Scalable Unsupervised Segmentation via Random Fourier Feature-based Gaussian Process

TL;DR

This paper introduces RFF-GP-HSMM, a scalable and efficient unsupervised time-series segmentation method that uses random Fourier features to significantly reduce computational costs while maintaining high performance.

Contribution

It proposes a novel RFF-based approximation for Gaussian process models, enabling fast segmentation of large-scale time-series data without matrix inversion.

Findings

Achieves segmentation performance comparable to traditional methods.

Runs approximately 278 times faster on large datasets.

Effectively handles high-dimensional, long time-series data.

Abstract

In this paper, we propose RFF-GP-HSMM, a fast unsupervised time-series segmentation method that incorporates random Fourier features (RFF) to address the high computational cost of the Gaussian process hidden semi-Markov model (GP-HSMM). GP-HSMM models time-series data using Gaussian processes, requiring inversion of an N times N kernel matrix during training, where N is the number of data points. As the scale of the data increases, matrix inversion incurs a significant computational cost. To address this, the proposed method approximates the Gaussian process with linear regression using RFF, preserving expressive power while eliminating the need for inversion of the kernel matrix. Experiments on the Carnegie Mellon University (CMU) motion-capture dataset demonstrate that the proposed method achieves segmentation performance comparable to that of conventional methods, with approximately 278 times faster segmentation on time-series data comprising 39,200 frames.