Leon Ingelse, LASIGE researcher and former Master’s student, gave a presentation “Optimization of Feature Learning through Grammar-Guided Genetic Programming” at the Universidad Complutense de Madrid (UCM), on November 7, 2022, invited by Professor José Ignácio Hidalgo, where he presented the GeneticEngine tool as well as his Master’s work, supervised by Alcides Fonseca, LASIGE integrated member.

Feature Learning (FL) is key to well-performing machine learning models. However, the most popular FL methods lack interpretability, which is becoming a critical requirement of Machine Learning. Our researcher proposed two new FL methods build on top of our Grammar-Guided Genetic Programming (GGGP) framework Genetic Engine.

The first FL method leverages domain knowledge to (1) restrict the solution space of the GGGP algorithm to improve the performance, and (2) to improve the interpretability of the solutions. The second FL method incorporates an aggregation method to improve performance. Results are promising, showing significant improvements.