Multivariate LSTM-FCNs for Time Series Classification

By Fazle Karim, Somshubra Majumdar, Houshang Darabi, Samuel Harford. Neural Networks, 2018

1,048 citations recorded by the source corpus.

Over the past decade, multivariate time series classification has received great attention. We propose transforming the existing univariate time series classification models, the Long Short Term Memory Fully Convolutional Network (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN), into a multivariate time series classification model by augmenting the fully convolutional block with a squeeze-and-excitation block to further improve accuracy. Our proposed models outperform most state-of-the-art models while requiring minimum preprocessing. The proposed models work efficiently on various complex multivariate time series classification tasks such as activity recognition or action recognition. Furthermore, the proposed models are highly efficient at test time and small enough to deploy on memory constrained systems.

Subfields: Artificial Intelligence, Biomedical Engineering, Signal Processing

View Paper: https://www.semanticscholar.org/paper/11c295c4e1ea15d5abd0c79aac2231783d243bdf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Anthony J. Bagnall
2. James Large
3. Jessica Lin
4. Yifeng Gao
5. Matthew Middlehurst
6. S. Mei
7. Chang-Tien Lu
8. Trond Kvamsdal

In Biomedical Engineering

1. Trond Kvamsdal
2. James Large
3. Omer San
4. Anthony J. Bagnall
5. Fan Yang
6. Adil Rasheed

In Signal Processing

1. Anthony J. Bagnall
2. James Large
3. Jessica Lin
4. Yifeng Gao
5. Matthew Middlehurst
6. Chang-Tien Lu
7. Alejandro Pasos Ruiz
8. Trond Kvamsdal

Citations in the corpus

• The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances (2020, methodology): “Memory Convolutional Network for time series classification (LSTM-FCN) [10] is adapted for the multivariate case and the fully convolutional blocks augmented with squeeze-and-excitation blocks to form MLSTM-FCN [9].”
• TapNet: Multivariate Time Series Classification with Attentional Prototypical Network (2020, methodology): “These techniques often use LSTM or CNN to extract the features (Karim et al. 2018a)(Zhao et al. 2017).”
• Digital Twin: Values, Challenges and Enablers From a Modeling Perspective (2019, methodology): “Karim et al. [121] demonstrated the use of LSTM in timeseries classification.”
• Multivariate Temporal Convolutional Network: A Deep Neural Networks Approach for Multivariate Time Series Forecasting (2019, background): “For the MALSTM-FCN network, the optimal number of LSTM hidden states for each dataset was found via grid search over { 8, 50, 100, 200 } .”