/MTCN

Multi-Task Temporal Convolutional Network

Primary LanguagePythonApache License 2.0Apache-2.0

MTCN

Multi-Task Temporal Convolutional Network

This is the MTCN model for Multi-Task Temporal Convolutional Network for Predicting Water Quality Sensor Data accepted by ICONIP 2019. Paper_online: (https://link.springer.com/chapter/10.1007/978-3-030-36808-1_14)

Data set is not included in this repository. Users can test the model on any time series data.


Important

model_hpc.py is the file used to get the final model for the paper.

This code is used for training on HPC, and it has some dedicated codes for parallel GPU training.

model_run.py is used locally for testing and valication. It has the same model but some parameters may differ from the paper.

The paper use the best models I selected from HPC training

Required packages:

  • Tensorflow
  • tcn
  • sklearn
  • numpy
  • pandas
  • matplotlib