Titulo Estágio
Evolving Grammar Forests
Áreas de especialidade
Sistemas Inteligentes
Sistemas Inteligentes
Local do Estágio
CISUC
Enquadramento
Genetic Programming (GP) is an Artificial Intelligence method inspired by the Theory of Evolution of Charles Darwin, where evolving solutions encode algorithmic strategies, i.e. computer programs to solve a given problem. These solutions can be encoded using different types of representations, such as parse-trees or grammars. The latter has gained some relevance in recent years, mainly due to Grammatical Evolution (GE) [1,2]. GE allows practitioners to define a grammar that defines the syntactic structure of the solutions to the problem at hand, making it easier for practitioners to effortlessly insert domain knowledge within the search process.
The objective of this work is to design, implement and analyse a framework that allows for the evolution of Machine Learning classifiers, similar to Random Forests, using a grammar. The main goal is to access the ability of creating an Evolutionary Framework that is able to evolve effective classifiers.
1 - O'Neill, M., & Ryan, C. (2001). Grammatical evolution. IEEE Transactions on Evolutionary Computation, 5(4), 349-358.
2- Lourenço, N., Pereira, F. B., & Costa, E. (2016). Unveiling the properties of structured grammatical evolution. Genetic Programming and Evolvable Machines, 17(3), 251-289.
Objetivo
To design, implement and validate an evolutionary approach for the design of random forests. The experimental validation will be performed using benchmarks widely used in the literature, and we will compare the results with the state of the art.
Plano de Trabalhos - Semestre 1
1 - Review of literature .
2 - Familiarisation with Grammatical Evolution and its variants.
3 - Design and proposal of the first version of the framework
4 - Implementation of the first version of the framework
5 - Writing of the intermediate repor
Plano de Trabalhos - Semestre 2
6 - Analysis of the preliminary results.
7 - Refinement of the framework
8 - Experimental Study
9 - Scientific Article with the main results
10 - Writing of the thesis
Condições
This internship results from a collaboration between the ECOS group and the NOVA Information Management School (NOVA IMS) of the Universidade Nova de Lisboa. The work is to be conducted at the ECOS laboratories of CISUC, with occasional meetings in Lisbon, with all the expenses supported. A workplace will be provided as well as the required computational resources.
Observações
N/A
Orientador
Nuno Lourenço / Leonardo Vanneschi
naml@dei.uc.pt 📩