Using Quantitative Genetics and Phenotypic Traits in Genetic Programming

Exploring genetic programming systems with MAP-Elites

10.7287/peerj.preprints.27154 ◽

2018 ◽

Author(s):

Emily Dolson ◽

Alexander Lalejini ◽

Charles Ofria

Keyword(s):

Neural Networks ◽

Genetic Programming ◽

Pareto Front ◽

Underlying Structure ◽

Phenotypic Traits ◽

Computation Technique ◽

Phenotype Space ◽

Programming Systems ◽

Insight Into ◽

Connection Cost

MAP-Elites is an evolutionary computation technique that has proven valuable for exploring and illuminating the genotype-phenotype space of a computational problem. In MAP-Elites, a population is structured based on phenotypic traits of prospective solutions; each cell represents a distinct combination of traits and maintains only the most fit organism found with those traits. The resulting map of trait combinations allows the user to develop a better understanding of how each trait relates to fitness and how traits interact. While MAP-Elites has not been demonstrated to be competitive for identifying the optimal Pareto front, the insights it provides do allow users to better understand the underlying problem. In particular, MAP-Elites has provided insight into the underlying structure of problem representations, such as the value of connection cost or modularity to evolving neural networks. Here, we extend the use of MAP-Elites to examine genetic programming representations, using aspects of program architecture as traits to explore. We demonstrate that MAP-Elites can generate programs with a much wider range of architectures than other evolutionary algorithms do (even those that are highly successful at maintaining diversity), which is not surprising as this is the purpose of MAP-Elites. Ultimately, we propose that MAP-Elites is a useful tool for understanding why genetic programming representations succeed or fail and we suggest that it should be used to choose selection techniques and tune parameters.

Download Full-text

Challenges in estimating heritability of phase polyphenism: insights from measured and simulated data in the desert locust

10.1101/149542 ◽

2017 ◽

Author(s):

Hélène Jourdan-Pineau ◽

Benjamin Pélissié ◽

Elodie Chapuis ◽

Floriane Chardonnet ◽

Christine Pagès ◽

...

Keyword(s):

Maternal Effects ◽

Quantitative Genetics ◽

Schistocerca Gregaria ◽

Simulated Data ◽

High Rate ◽

Phenotypic Traits ◽

Desert Locust ◽

Experimental Dataset ◽

Very Large Datasets ◽

Variance Estimates

AbstractQuantitative genetics experiments aim at understanding and predicting the evolution of phenotypic traits. Running such experiments often bring the same questions: Should I bother with maternal effects? Could I estimate those effects? What is the best crossing scheme to obtain reliable estimates? Can I use molecular markers to spare time in the complex task of keeping track of the experimental pedigree?We explored those practical issues in the desert locust, Schistocerca gregaria using morphologic and coloration traits, known to be influenced by maternal effects. We ran quantitative genetic analyses with an experimental dataset and used simulations to explore i) the efficiency of animal models to accurately estimate both heritability and maternal effects, ii) the influence of crossing schemes on the precision of estimates and iii) the performance of a marker-based method compared to the pedigree-based method.The simulations indicated that maternal effects deeply affect heritability estimates and very large datasets are required to properly distinguish and estimate maternal effects and heritabilities. In particular, ignoring maternal effects in the animal model resulted in overestimation of heritabilities and a high rate of false positives whereas models specifying maternal variance suffer from lack of power. Maternal effects can be estimated more precisely than heritabilities but with low power. To obtain better estimates, bigger datasets are required and, in the presence of maternal effects, increasing the number of families over the number of offspring per families is recommended. Our simulations also showed that, in the desert locust, using relatedness based on available microsatellite markers may allow reasonably reliable estimates while rearing locusts in group.In the light of the simulation results, our experimental dataset suggested that maternal effects affected various phase traits. However the statistical limitations, revealed by the simulation approach, didn’t allow precise variance estimates. We stressed out that doing simulations is a useful step to design an experiment in quantitative genetics and interpret the outputs of the statistical models.

Download Full-text

Exploring genetic programming systems with MAP-Elites

10.7287/peerj.preprints.27154v1 ◽

2018 ◽

Cited By ~ 1

Author(s):

Emily Dolson ◽

Alexander Lalejini ◽

Charles Ofria

Keyword(s):

Neural Networks ◽

Genetic Programming ◽

Pareto Front ◽

Underlying Structure ◽

Phenotypic Traits ◽

Computation Technique ◽

Phenotype Space ◽

Programming Systems ◽

Insight Into ◽

Connection Cost

MAP-Elites is an evolutionary computation technique that has proven valuable for exploring and illuminating the genotype-phenotype space of a computational problem. In MAP-Elites, a population is structured based on phenotypic traits of prospective solutions; each cell represents a distinct combination of traits and maintains only the most fit organism found with those traits. The resulting map of trait combinations allows the user to develop a better understanding of how each trait relates to fitness and how traits interact. While MAP-Elites has not been demonstrated to be competitive for identifying the optimal Pareto front, the insights it provides do allow users to better understand the underlying problem. In particular, MAP-Elites has provided insight into the underlying structure of problem representations, such as the value of connection cost or modularity to evolving neural networks. Here, we extend the use of MAP-Elites to examine genetic programming representations, using aspects of program architecture as traits to explore. We demonstrate that MAP-Elites can generate programs with a much wider range of architectures than other evolutionary algorithms do (even those that are highly successful at maintaining diversity), which is not surprising as this is the purpose of MAP-Elites. Ultimately, we propose that MAP-Elites is a useful tool for understanding why genetic programming representations succeed or fail and we suggest that it should be used to choose selection techniques and tune parameters.

Download Full-text