On Mini-Batch Training with Varying Length Time Series

TL;DR

This paper introduces a novel length normalization method for variable-length time series using Dynamic Time Warping, improving neural network training by better preserving dataset features compared to traditional padding or truncation.

Contribution

The paper proposes a new length normalization technique leveraging DTW to maintain dataset features, outperforming existing methods in neural network training on variable-length time series.

Findings

DTW-based normalization outperforms 18 other methods

Improves CNN, LSTM, and BLSTM performance on variable-length data

Applicable to real-world time series datasets

Abstract

In real-world time series recognition applications, it is possible to have data with varying length patterns. However, when using artificial neural networks (ANN), it is standard practice to use fixed-sized mini-batches. To do this, time series data with varying lengths are typically normalized so that all the patterns are the same length. Normally, this is done using zero padding or truncation without much consideration. We propose a novel method of normalizing the lengths of the time series in a dataset by exploiting the dynamic matching ability of Dynamic Time Warping (DTW). In this way, the time series lengths in a dataset can be set to a fixed size while maintaining features typical to the dataset. In the experiments, all 11 datasets with varying length time series from the 2018 UCR Time Series Archive are used. We evaluate the proposed method by comparing it with 18 other length normalization methods on a Convolutional Neural Network (CNN), a Long-Short Term Memory network (LSTM), and a Bidirectional LSTM (BLSTM).