Neural-Symbolic Integration: explainability and reasoning in KENN

The project combined the insights gained on KENN

Micro-project in Humane-AI-Net

neural symbolic

Short description

KENN is an acronym for Knowledge Enhanced Neural Network, a novel knowledge fusion architecture.

This project builds on earlier work by FBK in Trento on KENN  and by VUA in Amsterdam and combined both insights. The project pursued 3 aims: 

  1. To develop versions of KENN based on new t-conorms (like the product t-conorm and Łukasiewicz).
  2. To try to extend the expressivity of the logical constraints in KENN from sets of clauses to implications.
  3. To make it possible to check the exact contribution of each clause to the final predictions of KENN. 

Results live up to these promises.

Additional information

Output

  • paper describing improvements to KENN, published in workshop or conference
  • software: new version of KENN

This Humane-AI-Net micro-project was carried out by Stichting VU (Frank van Harmelen and Fondazione Bruno Kessler (FBK, Luciano Serafini,).

Technical categories
Knowledge Representation Machine learning
Technology Readiness Level
TRL 1-2 (Basic research)

Last updated
Sat, 04/30/2022 - 15:55
Contact details
Frank van Harmelen (frank.van.harmelen@cs.vu.nl)

Assets related to Neural-Symbolic Integration: explainability and reasoning in KENN