Lucid—A Genetic Programming Library for Apache Spark

Download Slides

Many popular ML models are inherently black-box, leaving little room for interpretation and explanation of predictions or the model itself. This becomes problematic if one needs to be well-informed about the analyzed process, control it (e.g. manufacturing), or communicate the findings to third-parties (healthcare, insurance). Lucid is an ongoing research project, funded by the Polish National Centre for Research and Development, aiming at delivering tools for inducing transparent models from data. The knowledge representations used in Lucid are symbolic expressions and rules, which can be easily interpreted by humans and, at the same time, efficiently used for automatic classification or regression.
The presentation will begin with a short introduction to explanatory modelling. The following section will explain how Genetic Programming can be applied to create transparent models and how Lucid implements this paradigm. The session will conclude with the summary of recent developments in AI and legal frameworks that call for the use of explanatory modelling.

Session hashtag: #EUres4

About Jakub Guner

Jakub Guner is a 2017 graduate of Poznan University of Technology, Poland. He specializes in Machine Learning and his Master Thesis involved development of Genetic Programming library for Apache Spark. He interned at MasterCard and Microsoft.