A Study on Segmentation of Time Series Data Using Neural Process

Abstract

In order to understand time series data, it is necessary to divide the data into meaningful chunks. However, it is difficult to manually segment a large amount of data. In this study, we discuss a computational model for segmenting time series data using an unsupervised learning model. In this paper, we compare the accuracy of segmentation by applying the Neural Process, Attentive Neural Process, and Conditional Neural Process to a generative model using the Neural Processes and HDP-HSMM.