Demos for PSyKI

This repository is meant to present demos for PSyKI, a python package for symbolic knowledge injection (SKI). In our intentions, a user should be able to get started with PSyKI by cloning this repository and running the notebooks in the notebooks directory.

There exists a doker image for the demos!

docker pull pikalab/demo-psyki-python:latest

For M1 users

docker pull pikalab/demo-psyki-python:latest-apple-m1

To execute the container run

docker run -it --rm -p 8888:8888 pikalab/demo-psyki-python

Before starting

All notebooks are self-contained and should be runnable in a fresh environment. Make sure to have satisfied the requirements (see requirements.txt):

(for developers)

• build 0.10.0
• setuptools 67.6.0
• treon 0.1.4

(for users and developers)

• jupyter 1.0.0
• tensorflow 2.7.0
• psyki 0.3.10
• pandas 1.5.3
• scikit-learn 1.2.2

Demos

• KBANN, the algorithm is presented in the context of a biological task consisting in classifying DNA sequences. It makes use of propositional logic rules, see knowledge\splice-junction.pl. The demo is available at notebooks\kbann.ipynb;
• KINS, a dataset concerning the yearly income of people living in the U.S. is used to present the algorithm. KBANN accepts rules in stratified (a.k.a. no recursion) Datalog or subsets of s. Datalog (e.g., propositional logic), see see knowledge\census-income.pl. The demo is available at notebooks\kins.ipynb.