Temporal Convolutional Explorer Helps Understand 1D-CNN’s Learning Behavior in Time Series Classification from Frequency Domain

By Junru Zhang et al

Published on June 10, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Work
3. Preliminaries
4. Temporal Convolutional Explorer
4.1 FE Feature Maps
4.2 Learning Behavior & Disturbing Convolution
5. Conclusion

Summary

The document discusses the Temporal Convolutional Explorer (TCE) mechanism to understand 1D-CNN's learning behavior in time series classification. It explores the frequency domain perspective and introduces concepts like FE feature maps, focus scale, and frequency centroid to identify frequency components. The analysis reveals that deeper 1D-CNNs may contain disturbing convolutions leading to accuracy degradation. The work aims to improve the performance of deep 1D-CNNs for TSC tasks.