LSTM Fully Convolutional Networks for Time Series Classification

By Fazle Karim et al

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

Table of Contents

I. Introduction
II. Background Works
III. LSTM Fully Convolutional Network
A. Network Architecture
B. Network Input
C. Fine-Tuning of Models
IV. Experiments
A. Evaluation Metrics
B. Results

Summary

The document discusses LSTM Fully Convolutional Networks for Time Series Classification. It proposes models that enhance the performance of fully convolutional networks by incorporating LSTM RNN sub-modules. The models achieve state-of-the-art performance on time series datasets. They utilize attention mechanisms and fine-tuning to further improve classification. The experiments show significant improvements over existing models on UCR time series datasets.