Genetic Algorithms have recently been successfully applied to the Machine Learning framework, being able to train autonomous agents and proving to be valid alternatives to state-of-the-art Reinforcement Learning techniques. Their attractiveness relies on the simplicity of their formulation and stability of their procedure, making them an appealing choice for Machine Learning applications where the complexity and instability of Deep Reinforcement Learning techniques is still an issue. However, despite their apparent potential, the classic formulation of Genetic Algorithms is unable to solve Machine Learning problems in the presence of high variance of the fitness function, which is common in realistic applications.

Gradient Bias to Solve the Generalization Limit of Genetic Algorithms Through Hybridization with Reinforcement Learning

Espositi Federico;Bonarini Andrea
2021

Abstract

Genetic Algorithms have recently been successfully applied to the Machine Learning framework, being able to train autonomous agents and proving to be valid alternatives to state-of-the-art Reinforcement Learning techniques. Their attractiveness relies on the simplicity of their formulation and stability of their procedure, making them an appealing choice for Machine Learning applications where the complexity and instability of Deep Reinforcement Learning techniques is still an issue. However, despite their apparent potential, the classic formulation of Genetic Algorithms is unable to solve Machine Learning problems in the presence of high variance of the fitness function, which is common in realistic applications.
Machine Learning, Optimization, and Data Science
978-3-030-64583-0
Reinforcement Learning
Genetic Algorithms
File in questo prodotto:
File Dimensione Formato  
Gradient_Bias_to_solve_the_Generalization_Limit_of_Genetic_Algorithms_through_hybridization_with_Reinforcement_Learning__Camera_Version_.pdf

embargo fino al 08/01/2022

Descrizione: Articolo principale
: Post-Print (DRAFT o Author’s Accepted Manuscript-AAM)
Dimensione 4.56 MB
Formato Adobe PDF
4.56 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11311/1170894
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact