Titulo Estágio
Making sexual selection accessible in evolutionary computation practice
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
CMS - CISUC
Enquadramento
This dissertation takes place within the scope of the ongoing research at CISUC concerning non-canonical evolutionary computation models. In particular, it builds upon previous research on the incorporation of sexual selection in evolutionary computation. Although these approaches showed promise, and have the potential to be applicable to a wide range of problems, their adoption is currently limited. As such, the proposal focuses on two core objectives: (i) making these approaches accessible to a large number of researchers by including them in existing evolutionary frameworks; (ii) further testing and development of the approaches in order to ascertain their adequacy and advantages.
Objetivo
While natural selection is often synthesized as “survival of the fittest” the phenomenon is significantly more complex. From an evolutionary perspective survival is not sufficient, there is a need to reproduce in order to pass on the genes to the next generation. Moreover, reproduction is costly, especially for females, meaning that females must chose their mating partners carefully in order to maximize the evolutionary success of the offspring. Conversely, in many scenarios, males are forced to compete among themselves to gain access to females. The evolutionary pressure of these forces results in a wide range of phenomena, including ornamentation and runaway sexual selection, and is considered a key driving force of evolution. Interestingly, most evolutionary computation models focus exclusively on the “survival of the fittest”, thus ignoring significant aspects of evolution.
In the course of his Ph.D., António Leitão studied sexual selection in the scope of evolutionary computation, proposing and developing several computational models of female mate choice. Among these models, the Preferences as Ideal Mating Partners (PIMP) approach stands out as the one which is easier to apply to a wide range of problems without significant overheads. In simple terms, in theory PIMP is applicable to any problem provided that it is possible to estimate the difference between the genotypes or phenotypes of two individuals. Moreover, the application is mostly effortless.
For these reasons, the current proposal has two main objectives:
(i) Making PIMP accessible to a large number of researchers by including it in existing evolutionary frameworks;
(ii) Further testing and development in order to ascertain their adequacy and advantages.
Plano de Trabalhos - Semestre 1
• Review of the state of the art in Sexual Selection
• Familiarisation with the models developed at CISUC
• Survey the existing evolutionary computation frameworks and selection of a subset;
• Incorporation of PIMP in the selected frameworks
• Intermediate report and other documentation
Plano de Trabalhos - Semestre 2
• Application of PIMP to benchmark problems and analysis of the results;
• Refinement of the model
• Application to “real world” problems
• Dissemination of the tools
• Writing the dissertation and other documentation
Condições
The work is to be conducted at the CMS group of CISUC. A workplace will be provided as well as the required computational resources.
Observações
Supervisors:
Penousal Machado
Carlos M. Fonseca
Orientador
Penousal Machado
machado@dei.uc.pt 📩