TY - GEN
T1 - A Study of Fitness Landscapes for Neuroevolution
AU - Rodrigues, Nuno M.
AU - Silva, Sara
AU - Vanneschi, Leonardo
N1 - info:eu-repo/grantAgreement/FCT/9471 - RIDTI/PTDC%2FCCI-CIF%2F29877%2F2017/PT#
info:eu-repo/grantAgreement/FCT/3599-PPCDT/DSAIPA%2FDS%2F0113%2F2019/PT#
info:eu-repo/grantAgreement/FCT/3599-PPCDT/DSAIPA%2FDS%2F0022%2F2018/PT#
info:eu-repo/grantAgreement/FCT/3599-PPCDT/PTDC%2FCCI-INF%2F29168%2F2017/PT#
Rodrigues, N. M., Silva, S., & Vanneschi, L. (2020). A Study of Fitness Landscapes for Neuroevolution. In 2020 IEEE Congress on Evolutionary Computation, CEC 2020: Conference Proceedings [9185783] (2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CEC48606.2020.9185783
PY - 2020/7
Y1 - 2020/7
N2 - Fitness landscapes are a useful concept to study the dynamics of meta-heuristics. In the last two decades, they have been applied with success to estimate the optimization power of several types of evolutionary algorithms, including genetic algorithms and genetic programming. However, so far they have never been used to study the performance of machine learning algorithms on unseen data, and they have never been applied to neuroevolution. This paper aims at filling both these gaps, applying for the first time fitness landscapes to neuroevolution and using them to infer useful information about the predictive ability of the method. More specifically, we use a grammar-based approach to generate convolutional neural networks, and we study the dynamics of three different mutations to evolve them. To characterize fitness landscapes, we study autocorrelation and entropic measure of ruggedness. The results show that these measures are appropriate for estimating both the optimization power and the generalization ability of the considered neuroevolution configurations.
AB - Fitness landscapes are a useful concept to study the dynamics of meta-heuristics. In the last two decades, they have been applied with success to estimate the optimization power of several types of evolutionary algorithms, including genetic algorithms and genetic programming. However, so far they have never been used to study the performance of machine learning algorithms on unseen data, and they have never been applied to neuroevolution. This paper aims at filling both these gaps, applying for the first time fitness landscapes to neuroevolution and using them to infer useful information about the predictive ability of the method. More specifically, we use a grammar-based approach to generate convolutional neural networks, and we study the dynamics of three different mutations to evolve them. To characterize fitness landscapes, we study autocorrelation and entropic measure of ruggedness. The results show that these measures are appropriate for estimating both the optimization power and the generalization ability of the considered neuroevolution configurations.
KW - Autocorrelation
KW - Convolutional Neural Networks
KW - Entropic Measure of Ruggedness
KW - Fitness Landscapes
KW - Neuroevolution
UR - http://www.scopus.com/inward/record.url?scp=85092063390&partnerID=8YFLogxK
UR - https://www.webofscience.com/wos/woscc/full-record/WOS:000703998202042
U2 - 10.1109/CEC48606.2020.9185783
DO - 10.1109/CEC48606.2020.9185783
M3 - Conference contribution
AN - SCOPUS:85092063390
T3 - 2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings
BT - 2020 IEEE Congress on Evolutionary Computation, CEC 2020
PB - Institute of Electrical and Electronics Engineers (IEEE)
T2 - 2020 IEEE Congress on Evolutionary Computation, CEC 2020
Y2 - 19 July 2020 through 24 July 2020
ER -