Semantic-based Machine Learning Genetic Programming

Abstract

The goal of having computers automatically solve problems is at the very core of artificial intelligence. Genetic Programming (GP) is an extremely successful Machine Learning Evolutionary Algorithm (EA) that has been used to e.g., optimise source code, automatically create controllers for games, classify highly imbalanced data, optimise energy usage. GP *automatically* solves problems starting from a high-level statement of what needs to be done without requiring the user to specify the form of the solution in advance. This talk covers basic concepts that help to understand how EAs work. We then focus our attention on the powerful representation used in GP. Next, we move to explain one of the hottest topics in GP: Semantics, and how this is making an important positive impact in GP search. We finalise this talk by discussing how semantics can be used in Evolutionary Multi-objective optimisation, one of the most popular research areas in Computer Science given e.g., its successful broad-applicability.

Short Bio

Edgar Galvan is an ELEVATE Fellow, funded by the Irish Research Council and co-funded by Marie Curie Actions. Edgar is officially affiliated to TCD SCSS and he is currently in his outgoing phase in INRIA, Paris, France. His research interests are in EAs and Monte Carlo Tree Search, with applicability in real-world problems (e.g., data analytics, energy-based problems). He holds a BSc in Informatics and a MSc by Research in Artificial Intelligence both from University of Veracruz, Mexico and a PhD in EAs from the University of Essex, United Kingdom (2009). He has independently been ranked as one of the top 3% researchers in Genetic Programming. He is constantly involved in (co-) organising workshops in international EAs conferences and special issues in EA journals.

Venue

Large Conference Room, O'Reilly Institute