DBA for Time Series Signal Classification

This is the official implementation of Dynamic Bilinear Low-Rank Attention (DBA) on time series signal classification.

Citation

If this repository is helpful to your research, we'd really appreciate it if you could cite the following paper:

@ARTICLE{10843139,
  author={Qin, Bosheng and Li, Juncheng and Tang, Siliang and Zhuang, Yueting},
  journal={IEEE Transactions on Neural Networks and Learning Systems}, 
  title={DBA: Efficient Transformer With Dynamic Bilinear Low-Rank Attention}, 
  year={2025},
  volume={},
  number={},
  pages={1-15},
  keywords={Transformers;Complexity theory;Attention mechanisms;Memory management;Training;Kernel;Sparse matrices;Optimization;Learning systems;Image coding;Bilinear optimization;dynamic compression;efficient transformer;low-rank attention},
  doi={10.1109/TNNLS.2025.3527046},
  ISSN={2162-2388},
  month={},}

Requirements

Hardware

1 Nvidia GPU.

Software

pip install -r requirements.txt
mkdir ./experiments
mkdir ./dataset

Dataset

Download the UEA multivariate time series classification archive and place it in the root folder as follows:

|-- dataset
	|-- Multivariate2018_ts
	|  |-- Multivariate_ts
	   |  |-- EthanolConcentration
           |  |-- FaceDetection
           ...

Experiments

Ethanolconcentration

bash EthanolConcentration.sh

Facedetection

bash FaceDetection.sh

Handwriting

bash handwriting.sh

Heartbeat

bash heartbeat.sh

Japanese vowels

bash japanese.sh

Pems-Sf

bash pems.sh

Selfregulationscp1

Selfregulationscp1.sh

Selfregulationscp2

bash Selfregulationscp2.sh

Spokenarabicdigits

bash SpokenArabicDigits.sh

Uwavegesturelibrary

UWaveGestureLibrary.sh