When a Genetic Algorithm (GA), or in general a stochastic algorithm, is employed in a statistical problem, the result is affected by both the variability due to sampling error, due to the fact that only a sample is observed, and the variability due to the stochastic nature of the algorithm. Such issues can be analyzed by understanding the trade-off between statistical accuracy and computational efforts. We focus on statistical estimation problems for which the variability of the GA estimates can be decomposed in the two sources of variability by means of cost functions, related to both data acquisition and runtime of the algorithm. Simulation studies will be presented to discuss the statistical and computational tradeoff question.
Statistical and computational tradeoff in econometric models building by genetic algorithms / Rizzo, Manuel; Battaglia, Francesco. - STAMPA. - (2016), pp. 89-89. (Intervento presentato al convegno 10th International Conference on Computational and Financial Econometrics (CFE 2016) and 9th International Conference of the ERCIM (European Research Consortium for Informatics and Mathematics) Working Group on Computational and Methodological Statistics (CMStatistics 2016) tenutosi a Sevilla, Spain nel December 9th - December 11th).
Statistical and computational tradeoff in econometric models building by genetic algorithms
RIZZO, MANUEL;BATTAGLIA, Francesco
2016
Abstract
When a Genetic Algorithm (GA), or in general a stochastic algorithm, is employed in a statistical problem, the result is affected by both the variability due to sampling error, due to the fact that only a sample is observed, and the variability due to the stochastic nature of the algorithm. Such issues can be analyzed by understanding the trade-off between statistical accuracy and computational efforts. We focus on statistical estimation problems for which the variability of the GA estimates can be decomposed in the two sources of variability by means of cost functions, related to both data acquisition and runtime of the algorithm. Simulation studies will be presented to discuss the statistical and computational tradeoff question.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
