What is it about?
Over recent years, there has been a rapid development of deep learning (DL) in both industry and academia fields. However, finding the optimal hyperparameters of a DL model often needs high computational cost and human expertise. To mitigate the above issue, evolutionary computation (EC) as a powerful heuristic search approach has shown significant merits in the automated design of DL models, so-called evolutionary deep learning (EDL). This paper aims to analyze EDL from the perspective of automated machine learning (AutoML). Specifically, we firstly illuminate EDL from DL and EC and regard EDL as an optimization problem. According to the DL pipeline, we systematically introduce EDL methods ranging from data preparation, model generation, to model deployment with a new taxonomy (i.e., what and how to evolve/optimize), and focus on the discussions of solution representation and search paradigm in handling the optimization problem by EC. Finally, key applications, open issues and potentially promising lines of future research are suggested. This survey has reviewed recent developments of EDL and offers insightful guidelines for the development of EDL.
Featured Image
Photo by Liam Charmer on Unsplash
Why is it important?
Evolutionary computation (EC) has been widely applied to automatic DL, termed as evolutionarydeep learning (EDL), owing to its flexibility and automatically evolving mechanism. In principle, EC evolves a population of individuals towards the optimum/optima via evolutionaryoperations and environmental selection, which is a gradient-free and robust search paradigm. As a result, the integration of EC and DL has become a hot research topic in both academic and industrialcommunities. Moreover, the number of publications and citations referring to EC & DL byyears from Web of Science gradually increases until around 2018, whereas it sharply rises in the recentyears. Hence, more and more researchers work on the area of EDL..
Perspectives
Existing work on EDL is reviewed from the perspective of DL and EC to facilitate the understandingof readers from the communities of both ML and EC, and we also formulated EDL into anoptimization problem from the perspective of EC. The survey describes and discusses on EDL in terms of data preparation, model generation and modeldeployment from a novel taxonomy, where the solution representation and the search paradigms are emphasized and systematically discussed. To the best of our knowledge, few survey has investigatedthe evolutionary model deployment. On the basis of the comprehensive review of EDL approaches, a number of applications, open issuesand trends of EDL are discussed, which will guide the development of EDL
Nan Li
Read the Original
This page is a summary of: Survey on Evolutionary Deep Learning: Principles, Algorithms, Applications and Open Issues, ACM Computing Surveys, June 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3603704.
You can read the full text:
Contributors
The following have contributed to this page
