Skip to content

"Interactive Polar Diagrams for Model Comparison" by Aleksandar Anžel, Dominik Heider, and Georges Hattab

License

Notifications You must be signed in to change notification settings

AAnzel/Polar-Diagrams-for-Model-Comparison

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Polar Diagrams for Model Comparison

GitHub Test Status GitHub Build Status GitHub Publish GitHub Issues PyPI - License PyPI PyPI - Status PyPI - Downloads

Manuscript

This library is created for the following paper:

"Interactive Polar Diagrams for Model Comparison" by Aleksandar Anžel, Dominik Heider, and Georges Hattab

Please cite the paper as:

@article{ANZEL2023107843,
title = {Interactive polar diagrams for model comparison},
journal = {Computer Methods and Programs in Biomedicine},
volume = {242},
pages = {107843},
year = {2023},
issn = {0169-2607},
doi = {https://doi.org/10.1016/j.cmpb.2023.107843},
url = {https://www.sciencedirect.com/science/article/pii/S0169260723005096},
author = {Aleksandar Anžel and Dominik Heider and Georges Hattab},
keywords = {Bioinformatics, Machine-learning, Visualization, Evaluation, Climate, Comparison, Ai, Data-visualization, Information-visualization, Predictive-analysis, Model-comparison, Climate-model-visualization, Ml-model-evaluation, Taylor-diagram, Mutual-information-diagram, Entropy, Mutual-information, Variation-of-information, Correlation, Medical-data},
abstract = {Objective
Evaluating the performance of multiple complex models, such as those found in biology, medicine, climatology, and machine learning, using conventional approaches is often challenging when using various evaluation metrics simultaneously. The traditional approach, which relies on presenting multi-model evaluation scores in the table, presents an obstacle when determining the similarities between the models and the order of performance.
Methods
By combining statistics, information theory, and data visualization, juxtaposed Taylor and Mutual Information Diagrams permit users to track and summarize the performance of one model or a collection of different models. To uncover linear and nonlinear relationships between models, users may visualize one or both charts.
Results
Our library presents the first publicly available implementation of the Mutual Information Diagram and its new interactive capabilities, as well as the first publicly available implementation of an interactive Taylor Diagram. Extensions have been implemented so that both diagrams can display temporality, multimodality, and multivariate data sets, and feature one scalar model property such as uncertainty. Our library, named polar-diagrams, supports both continuous and categorical attributes.
Conclusion
The library can be used to quickly and easily assess the performances of complex models, such as those found in machine learning, climate, or biomedical domains.}
}

DOI


Abstract:

Objective: Evaluating the performance of multiple complex models, such as those found in biology, medicine, climatology, and machine learning, using conventional approaches is often challenging when using various evaluation metrics simultaneously. The traditional approach, which relies on presenting multi-model evaluation scores in the table, presents an obstacle when determining the similarities between the models and the order of performance.

Methods: By combining statistics, information theory, and data visualization, juxtaposed Taylor and Mutual Information Diagrams permit users to track and summarize the performance of one model or a collection of different models. To uncover linear and nonlinear relationships between models, users may visualize one or both charts.

Results: Our library presents the first publicly available implementation of the Mutual Information Diagram and its new interactive capabilities, as well as the first publicly available implementation of an interactive Taylor Diagram. Extensions have been implemented so that both diagrams can display temporality, multimodality, and multivariate data sets, and feature one scalar model property such as uncertainty. Our library, named \emph{polar-diagrams}, supports both continuous and categorical attributes.

Conclusion: The library can be used to quickly and easily assess the performances of complex models, such as those found in machine learning, climate, or biomedical domains.

Dependencies

The code is written in Python 3.9.15 and tested on Linux with the following libraries installed:

Library Version
numpy 1.23.5
pandas 1.5.2
scikit-learn 1.2.0
scipy 1.9.3
plotly 5.9.0
kaleido 0.2.1

The dependencies can also be found in requirements.txt.

Data

Location Description
Data/ contains all datasets used in Source/main.ipynb.
Data/Dataset_0/ contains the Anscombe's quartet data set and the Datasaurus data set.
Data/Dataset_1/ contains the official, automatically generated script for downloading the CMIP3 data from the https://esgf-node.llnl.gov/projects/cmip3/. To generate the whole data set, the user should first place itself into this directory and then run the following command from the terminal sh generate_dataset_1.sh. [1]
Data/Dataset_2/ contains the data set from the subsection 3.2 Example 2 — Machine Learning Model Evaluation of our paper.
Data/Dataset_3/ contains the data set from the subsection 3.3 Example 3 — Biomedical Similarity Assertion of our paper.

[1] The script used for downloading the Dataset_1/ was generated using the tutorial found here https://esgf.github.io/esgf-user-support/faq.html#how-to-preserve-the-directory-structure. Script can be automatically generated and downloaded again from here https://esgf-data.dkrz.de/esg-search/wget?download_structure=model&project=CMIP3&experiment=historical&ensemble=run1&variable=ts.

Code

Source Code Description
Source/ contains all source scripts.
Source/main.ipynb contains the IPython (jupyter) notebook that demonstrates the library using multiple datasets. This notebook reproduces all of the results we presented in our paper.
Source/polar_diagrams/ contains the root source code directory of our library.
Source/polar_diagrams/src/polar_diagrams/polar_diagrams.py contains the source code that imports the data, modifies it, calculates statistical and information theory properties, and builds diagrams.
Source/polar_diagrams/tests/test.py contains the source code for all unit tests.

Notable Features

Scalar Feature

Scalar feature

Two-version Model Feature

Two-versions feature

Interactive Aspects

Tooltip

Installation

Stable

We recommend installing the library using pip:

pip install polar-diagrams

Unstable

git clone https://github.com/AAnzel/Polar-Diagrams-for-Model-Comparison.git
cd Polar-Diagrams-for-Model-Comparison/Source/polar_diagrams/
pip install --editable .

Running

Please check the API documentation of our library at Source/polar_diagrams/docs/polar_diagrams.md or the IPython (jupyter) notebook that demonstrates the library at Source/main.ipynb.

License

Licensed under the GNU General Public License, Version 3.0 (LICENSE or https://www.gnu.org/licenses/gpl-3.0.en.html)

Contribution

Any contribution intentionally submitted for inclusion in the work by you, shall be licensed under the GNU GPLv3.