Skip to content

emanuelsla/HybridXIML

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Hybrid Explanatory Interactive Machine Learning

This repository contains supplementary code to our paper with its full reference:


Slany, Emanuel, Stephan Scheele, and Ute Schmid (2024). “Hybrid Explanatory Interactive Machine Learning for Medical Diagnosis”. In: Artificial Intelligence Applications and Innovations. Ed. by Ilias Maglogiannis et al. Cham: Springer Nature Switzerland, pp. 105–116. DOI: 10.1007/978-3-031-63211-2_9.


In sensitive domains such as medicine, machine learning faces two cruical requirements: Experts - in this case physicians - must be able to influence the model during the optimization. And it must be guaranteed that the experts' feedback persists for similar instances. Our method satisfies both requirements by combining the state-of-the-art CAIPI algorithm with probabilistic logic rule learning.

Quick start

$ pip install -r requirements.txt

($ pip uninstall certifi, module not required)

$ python3 -m scripts.main

Executing the prior commands install the project's requirements and start a hyXIML optimization cycle. Hyper-parameter descriptions can be seen by:

$ python3 -m scripts.main -h

Reproduction of paper experiments

Our paper experiments can be reproduced as follows:

$ python3 -m scripts.main

$ python3 -m scripts.generate_theories

$ python3 -m scripts.caipi_hximl_comparison

Please use the hyper-parameters from the referenced article. The first command initiates the logging directories, the second generates the theories (if not already done by the first), and the final command compares ML to logical inferences - or, CAIPI to hyXIML.

Finally, the command

$ python3 -m scripts.apply_hximl_from_logging

executes hyXIML directly from the logging directory, using existing models and theories.

About

Code to paper: Emanuel Slany, Stephan Scheele, and Ute Schmid (2024): Hybrid Explanatory Interactive Machine Learning for Medical Diagnosis.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages