Event section image

A Memetic Algorithm for Symbolic Regression



Prof. Pablo Moscato

Prof. Pablo Moscato

School of Electrical Engineering and Computing
The University of Newcastle


1100-1200, 18 Jun 2019


SCT909, Cha Chi Ming Science Tower, Ho Sin Hang Campus


The talk presents a quick overview of the field of memetic algorithms which has now several thousand papers around the world in many different fields. A memetic algorithm, is among the most useful metaheuristics for complex optimisation problems. This paradigm of meta-heuristic optimization is based on the evolution of solutions by a set of computational agents. Feasible solutions are generated based on a representation that may also include non-feasible configurations. A population of computational agents with problem domain knowledge improves the set of feasible solutions using local search heuristics. In addition, the agents compete in searching for function models with fewer number of variables. Agent interactions are constrained by a population structure which has been previously used in several successful MAs for other combinatorial optimization problems.

In particular, Prof. Moscato will present work with Haoyuan Sun, from the California Institute of Technology, developed during 2018. Together the team addressed the practical difficulties of computational heuristics for symbolic regression, which models data with algebraic expressions. Moscato and Sun proposed a new representation since they argue that sometimes the target unknown function may be best represented as the ratio of functions. They introduce a new alternative, a quite general approach based on a different representation of mathematical models with an analytic continued fraction representation, from which rational function models can be extracted. They present results using a tree-based population structure to improve the algorithm’s consistency and performance. Data from real-world applications are used to measure the potential of the approach and benchmark its performance against other approaches for symbolic regression.


Pablo Moscato obtained his B.S. in Physics from National University of La Plata in 1987, Argentina, before completing a PhD at the University of Campinas, Brazil, in 2001. In 1988, while at the California Institute of Technology, he developed with M.G. Norman the first memetic algorithm for the Traveling Salesman Problem. He has championed the field of memetic computing since then. Over the past three decades, he introduced several new computer science methods that lead to advances in personalised medicine, including the discovery of biomarkers and hallmarks of cancer, Alzheimer and other human diseases.

Due to his lifetime achievements in interdisciplinary research and in memetic computing in particular, he was nominated to the Rotary STAR (Science, Technology, Aerospace, Robotics) 2018 Awards in the categories of “Health and Medical” and “Knowledge Sharing”. These awards are annually given “in recognition to outstanding scientific and technological achievements with significant humanitarian benefit”.

He has successfully supervised 19 PhD candidates to completion since 2002 and he is currently supervising another 5 PhD candidates. He has also gained more than $14 million dollars in project grant funding that helped to support 49 research projects over the past 17 years.

He has 275 scientific publications with four co-edited books including “New Ideas in Optimization” (199), “Handbook of Memetic Algorihtms” (2012) and “Business and Consumer Analytics: New Ideas” (released by Springer on May 31st, 2019).

He was an Australian Research Council (ARC) Future Fellow (2012-2016) and the Founding Director of the University of Newcastle’s Priority Research Centre (PRC) for Bioinformatics, Biomarker Discovery and Information-Based Medicine (2007-2015) and the Newcastle Bioinformatics Initiative (2003-2006). He is currently a Professor of Computer Science at The University of Newcastle, Australia.