antakia

Welcome to AntakIA !

Hi!

AntakIA is THE tool to explain an ML model or replace it with a collection of basic explainable models.

AntakIA can be used for many purposes: it may help you understand the truths your ML model has learned, or it can assist you in converting an ML model into multiple simple rule-based algorithms.

It achieves this by dividing a given ML model into smaller regions. Hence its name: Antakia is an ancient city famous for its mosaics (and IA is AI in French).

Here is a quick overview on AntakIA:

See full video on Youtube.

Install

AntakIA is available on PyPI :

pip install antakia

Running example notebooks

Once you've installed antakia, you can download some of our notebook examples from our repo here.

Then, launch a Jupyter server from the notebook file (.ipynb) location :

jupyter notebook # or jupyter lab

You'll find a complete tutorial for our California housing example here : https://doc.antakia.ai

[!IMPORTANT] If you're using a virtual env, it's handy to be able to chose it from Jupyter's kernel list. Do the following :

python -m ipykernel install --user --name your_venv_name --display-name "My great virtual env"

Online demo

You can give antakia a try online : https://demo.antakia.ai

Log as demo with password antakia. Note it runs on a simple server and may be busy.

Run with Docker

[!IMPORTANT] Be sure to have a Docker engine running on your computer (ie. launch Docker Desktop)

docker build -t antakia .
docker run -p 8888:8888 antakia

In your Terminal, click on the http://127.0.0.1:8888/lab?token=WHATEVER_YOUR_TOKEN_IS URL link.

Logs

antakia is our open-source project. We plan to learn a lot from it to build our SaaS product, and earn our living. By default, antakia sends anonymous usage logs to our server. With this data, we cannot discover who you are, what you do, or what data you work on. It only tells us how people use the software. However, if you don't want us to receive this data, just create a .env file at the root of your project folder with the following line SEND_ANONYMOUS_LOGS=false