DeepIOStability

Requirements

• python3 (> 3.3)
  • Pytorch
  • joblib
  • scikit-learn
  • scipy
  • optuna

conda install pytorch
conda install -c conda-forge optuna

Anaconda install

First, please install anaconda by the official anaconda instruction [https://conda.io/docs/user-guide/install/linux.html].

Installation of DeepIOStability

pip install git+https://github.com/clinfo/DeepIOStability.git

Command

The dios command for training

dios train --config <config file>

The dios command for plotting, where the plot signals are selected from validation data at random by default.

dios-plot --config <config file>

Demo

Making a sample dataset

$ cd sample02
$ python make.py

Execution

$ dios train --config config.json

linear model comparative method

$ dios-linear train --config config.json --method <method>

<method> is selected from

• ORT
• MOESP
• ORT_auto
• MOESP_auto

Full experiments

Our experimental results can be reproduced by executing the script in the directory experiments/.

To make a command list, please run the following commands:

sh all4parallel.sh

Please execute the all commands in the two command lists:

• all4parallel_gpu.list for GPU machines
• all4parallel_cpu.list for CPU machines

Hyperparameter tuning

dios-opt --config <config.json>

Using the Optuna library, this command execute a hypaparameter optimization based on a given config files.

Our configuration files, obtained as a result of our hyperparameter tuning, put in experiments/***/config001000.json.