DiversifyAnalysis

Overview

This repo reproduces the experiments from paper "OUT-OF-DISTRIBUTION REPRESENTATION LEARNING FOR TIME SERIES CLASSIFICATION"

How-to

1. preprocess

Download and prepare your datasets by:

python prepare_dataset.py --data_dir ./data/ --dataset EMG

if executed successfully, dataset EMG will be saved to ./data/emg

2. train

python train.py --data_dir ./data/ --task cross_people --test_envs 0 --dataset emg --algorithm diversify --latent_domain_num 10 --alpha1 1.0 --alpha 1.0 --lam 0.0 --local_epoch 3 --max_epoch 5 --lr 0.01 --output ./data/train_output/act/cross_people-emg-Diversify-0-10-1-1-0-3-50-0.01 --output_model ./data/model_output

• specify your dataset with --data_dir
• specify your dataset name with --dataset
• specify a different test group with --test_envs if appliable
• training log will be saved to --output
• generated models will be saved to --output_model
• see other parameters list with --help

3. evaluate

for each training round, the output value 'target acc' is the accuracy value for the model

References

@inproceedings{lu2022out,
  title={Out-of-distribution Representation Learning for Time Series Classification},
  author={Lu, Wang and Wang, Jindong and Sun, Xinwei and Chen, Yiqiang and Xie, Xing},
  booktitle={International Conference on Learning Representations},
  year={2023}
}