DOCSEARCHbyAlgolia
Submit Docs

ModernTCN Documentation

repository·main·Indexed 19 days ago

https://github.com/luodhhh/moderntcn

A pure convolution structure for general time series analysis, supporting forecasting (long-term and short-term), imputation, classification, and anomaly detection.

Tokens
501
Snippets
6
Records
6
Agent score
16%

What's inside ModernTCN

  1. Run Short-term forecasting experiments

    main

    Short-term forecasting experiment code is located in the ./ModernTCN-short-term directory. To run the experiment on the M4 dataset, navigate to the short-term directory and execute the M4 shell script.

    cd ./ModernTCN-short-term
sh ./scripts/M4.sh

  2. Run Long-term forecasting experiments

    main

    Long-term forecasting experiment code is located in the ./ModernTCN-Long-term-forecasting directory. Experiment scripts are stored in the ./scripts subfolder. To run the experiment on the ETTh2 dataset, navigate to the forecasting directory and execute the ETTh2 shell script.

    cd ./ModernTCN-Long-term-forecasting
sh ./scripts/ETTh2.sh

  3. Run Imputation experiments

    main

    Imputation experiment code is located in the ./ModernTCN-imputation directory. To run the experiment on the ETTh2 dataset, navigate to the imputation directory and execute the ETTh2 shell script.

    cd ./ModernTCN-imputation
sh ./scripts/ETTh2.sh

  4. Run Classification experiments

    main

    Classification experiment code is located in the ./ModernTCN-classification directory. To run the classification experiments, navigate to the classification directory and execute the classification.sh script.

    cd ./ModernTCN-classification
sh ./scripts/classification.sh

  5. Install ModernTCN and dependencies

    main

    To use ModernTCN, ensure you have Python 3.7 installed. You can install the necessary dependencies using pip from the requirements.txt file provided in the repository.

    pip install -r requirements.txt

  6. Run Anomaly detection experiments

    main

    Anomaly detection experiment code is located in the ./ModernTCN-detection directory. To run the experiment on the SWaT dataset, navigate to the detection directory and execute the SWaT shell script.

    cd ./ModernTCN-detection
sh ./scripts/SWaT.sh