The definition and consolidation of Geometric Semantic Genetic Programming (GSGP), a family of novel Machine Learning algorithms based on Genetic Programming that, for the first time, can induce an unimodal error surface (i.e. with no local optima) for any supervised learning problem.
GSGP includes systems for regression, classification and forecasting and it was applied with success to several real-world domains, including the prediction of the relative position of computer tomography slices, of the proteins tertiary structure, of the Amazon’s products reviews, of energy consumption, of pharmacokinetic parameters, of the unified Parkinson’s disease rating scale, of high-performance concrete strength, of anticoagulation levels, and of forest aboveground biomass.
GSGP was the basis of an FCT project, on maritime safety and security, funded and concluded during the period: MaSSGP - Improving Semantic Genetic Programming for Maritime Safety, Security and Environmental Protection. These applications yield several publications in international journals, such as IEEE Transactions on Cybernetics, Expert Systems with Applications or Neurocomputing.
Following this line of work, Prof. L. Vanneschi received, in 2015, the Evostar Award for Outstanding Contributions in Evolutionary Computation, and also the European Conference on Genetic Programming Best Paper Award in 2013 and 2014.
Project: MaSSGP - Improving Semantic Genetic Programming for Maritime Safety, Security and Environmental Protection (2013-2015). Funded by Fundação para a Ciência e a Tecnologia