Omni-Scale CNNs: a simple and effective kernel size configuration for time series classification

TL;DR

This paper introduces an Omni-Scale block for 1D-CNNs that uses a simple rule to select kernel sizes, enabling models to effectively capture optimal receptive fields across diverse time series datasets and achieve state-of-the-art results.

Contribution

The paper proposes a universal Omni-Scale block with a simple kernel size rule, improving time series classification performance without dataset-specific tuning.

Findings

Achieves state-of-the-art performance on four benchmarks

Models with OS-block match performance of optimally tuned RF models

Universal kernel size rule effectively captures optimal RF across datasets

Abstract

The Receptive Field (RF) size has been one of the most important factors for One Dimensional Convolutional Neural Networks (1D-CNNs) on time series classification tasks. Large efforts have been taken to choose the appropriate size because it has a huge influence on the performance and differs significantly for each dataset. In this paper, we propose an Omni-Scale block (OS-block) for 1D-CNNs, where the kernel sizes are decided by a simple and universal rule. Particularly, it is a set of kernel sizes that can efficiently cover the best RF size across different datasets via consisting of multiple prime numbers according to the length of the time series. The experiment result shows that models with the OS-block can achieve a similar performance as models with the searched optimal RF size and due to the strong optimal RF size capture ability, simple 1D-CNN models with OS-block achieves the state-of-the-art performance on four time series benchmarks, including both univariate and multivariate data from multiple domains. Comprehensive analysis and discussions shed light on why the OS-block can capture optimal RF sizes across different datasets. Code available [https://github.com/Wensi-Tang/OS-CNN]