Bridging Logic Programming and Deep Learning for Explainability through ILASP
Talissa Dreossi
TL;DR
This work presents a hybrid AI framework that merges deep learning with Inductive Logic Programming to achieve accurate predictions accompanied by explanations. By leveraging ILASP and FastLAS for symbolic reasoning and XASP for explainability, the approach targets domains including weather forecasting, legal reasoning, and image recognition, aiming to retain predictive power while delivering human-understandable rules. Early results show promise: neural models provide forecasts and detections, while ILP-based methods offer transparent justifications and post-hoc explanations, with ongoing efforts to handle noise and vagueness and to scale computations via GPU acceleration. The work advances trustworthy AI by bridging high-performance learning with symbolic, interpretable reasoning, potentially enabling end-users to validate and understand AI-driven decisions in critical real-world contexts.
Abstract
My research explores integrating deep learning and logic programming to set the basis for a new generation of AI systems. By combining neural networks with Inductive Logic Programming (ILP), the goal is to construct systems that make accurate predictions and generate comprehensible rules to validate these predictions. Deep learning models process and analyze complex data, while ILP techniques derive logical rules to prove the network's conclusions. Explainable AI methods, like eXplainable Answer Set Programming (XASP), elucidate the reasoning behind these rules and decisions. The focus is on applying ILP frameworks, specifically ILASP and FastLAS, to enhance explainability in various domains. My test cases span weather prediction, the legal field, and image recognition. In weather forecasting, the system will predict events and provides explanations using FastLAS, with plans to integrate recurrent neural networks in the future. In the legal domain, the research focuses on interpreting vague decisions and assisting legal professionals by encoding Italian legal articles and learning reasoning patterns from Court of Cassation decisions using ILASP. For biological laboratories, we will collaborate with a research group to automate spermatozoa morphology classification for Bull Breeding Soundness Evaluation using YOLO networks and ILP to explain classification outcomes. This hybrid approach aims to bridge the gap between the high performance of deep learning models and the transparency of symbolic reasoning, advancing AI by providing interpretable and trustworthy applications.