2024-07-12 –, For Loop (3.2)
In this talk, we update on the developments of the Sole.jl framework for symbolic AI. In the past year we have been implementing a few algorithms for learning symbolic models (e.g., rules, lists, trees, forests), and for analyzing, evaluating, and post-processing the learned models.
Symbolic AI provides transparent alternatives to standard black-box AI algorithms, and includes methods for rule extraction, data mining and decision tree/list learning. The field is underdeveloped, its potential still to discover, and a programming framework for it is needed.
Sole.jl proposes itself as the first programming framework specifically tailored for symbolic AI workflows, and it puts specific focus on a symbolic representation of knowledge, in the abstract form of syntax trees. At the moment, it is still at an early stage, but its vision includes symbolic algorithms for processing data, for learning rules, trees, lists, and ensembles, and for post-hoc processing the learned models and their knowledge.
In this talk, we overview the structure of Sole.jl, and update on some advancements made in the past year.
Additional resources:
1. A 10-hour course on Modal Symbolic Learning with Sole.jl: YouTube playlist, Slides and Jupyter Notebooks;
2. A technical motivation for Sole.jl was presented at JuliaCon2023
Finishing a PhD with focus on Symbolic Learning and Logic in AI!
Enthusiast about Julia!!