trusty-ai-sandbox

A sandbox repository for the Trusty AI team

Content

Counterfactual explanation with OptaPlanner

Code of the demo we presented at OptaPlanner week about the usages of OptaPlanner to produce CF explaination: video + code

End to end POC: tracing + monitoring + explainability

POC to demonstrate how to use Kogito to execute a DMN and add tracing, monitoring and explainability features

End to end POC instructions

Note: the whole POC has been implemented and merged with Kogito codebase so you can use this guide to Kogito and reproduce the same features

Explainability research

explainability folder contains initial experiments of different approaches to apply local explainability to a decision service

Front-end POC

Playground used to design and implement initial version of Audit UI. Code has been merged on Kogito codebase and can be found here

PMML notebook

Example of a Jupyter notebook that shows how to train a model and export it as PMML to be integrated then to DMN (link). The notebook has been created starting from risk-pmml-builder example.

StackOverflow POC

POC that applies multiple algorithms to classify StackOverflow questions (link)