Towards measuring predictability

In folder MutualInformationMeasure one can find our implemetation of Mutual information measure.

In folder ChiSquareMeasure one can find our implemetation of Chi square measure

In folder PearsonCorrelationMeasure one can find our implemetation of Pearson correlation measure

In the file example.py, you will find a small working example to evaluate the information content between input and output and after training a model (MLP or Non-stationary transformer), it is evaluated how much information the residuals share with the input.

Theory

For detailed explanations, please see Towards Measuring Predictability , in course of which this repository was developed.

Requirements

All the required packages can be installed using the following command:

conda create --name predictability
conda activate predictability
pip3 install -r requirements.txt

How to Train

With predictability_measure as the working directory execute the python script python example.py which shows an example of how one can use our code.

Datasets

We have provided a sample csv file including a sinusoid plus noise in the data folder.

Citation

If you find our work useful, please consider citing our paper:

@article{saleh2024towards,
  title = {Towards Measuring Predictability: To which extent datadriven approaches can extract deterministic relations from data exemplified with time series prediction and classification},
  author = {Saleh, Gholam Zadeh and Vaisakh Shaj and Gerhard Neumann and Tim Breitenbach},
  journal = {Transactions on Machine Learning Research},
  year = {2024},
  url = {https://openreview.net/forum?id=jZBAVFGUUo&noteId=LEMTDMLbq7}
}

Thank you for your support!