Skip to content

A repository of examples for tools that implement model explainability

Notifications You must be signed in to change notification settings

fuzzylabs/explainability-examples

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Explainability Examples

What is explainability?

Explainability tools break down model inferences into how each input parameter influenced the final result. This has two main uses:

  • Feedback on individual inferences. If a user is rejected for a mortgage what can they do to improve? Increase their credit score, up their deposit, etc...
  • Monitoring for systematic bias in the model, by gathering values across a set of inferences. We can verify that our model is making reasonable decisions, and that we're not making incorrect or unjustified decisions, for example rejecting mortgage applicants users based on their gender.

it's also notable that their is a tradeoff between accuracy and explainability of models, see the Algorithmic Choice and Code section of [this article](Algorithmic Choice and Code) for more details.

What is shap

Shap is an academic python library that uses a game theory approach to graph model explainability. This article gives a breakdown of how specifically this is done.

Articles

Simplifying the Role of Explainability in the MLOps Cycle

Shap values explained

Installation and running

Run pip install -r requirements.txt to install project dependencies. Run python main.py to train an iris model and update the saved explainability graphs.

A force plot is saved to shap.html and a summary plot is saved to shap.png.

About

A repository of examples for tools that implement model explainability

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published