Lights and shadows in Evolutionary Deep Learning: Taxonomy, critical methodological analysis, cases of study, learned lessons, recommendations and challenges

摘要: Abstract Much has been said about the fusion of bio-inspired optimization algorithms and Deep Learning models for several purposes: from discovery network topologies hyperparametric configurations with improved performance a given task, to model’s parameters as replacement gradient-based solvers. Indeed, literature is rich in proposals showcasing application assorted nature-inspired approaches these tasks. In this work we comprehensively review critically examine contributions made so far based on three axes, each addressing fundamental question research avenue: (a) taxonomy (Why?), including historical perspective, definitions problems Learning, associated an in-depth analysis literature, (b) critical methodological (How?), which together two case studies, allows us address learned lessons recommendations good practices following (c) challenges new directions (What can be done, what for?). summary, axes – taxonomy, analysis, outline complete vision merger technologies drawing up exciting future area research.