{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/limited-evaluation-evolutionary-optimization","title":"Limited Evaluation Evolutionary Optimization of Large Neural Networks","arxiv_id":"1806.09819","date":"2018-06-26","proceeding":null,"authors":["Jonas Prellberg","Oliver Kramer"],"abstract":"Stochastic gradient descent is the most prevalent algorithm to train neural\nnetworks. However, other approaches such as evolutionary algorithms are also\napplicable to this task. Evolutionary algorithms bring unique trade-offs that\nare worth exploring, but computational demands have so far restricted\nexploration to small networks with few parameters. We implement an evolutionary\nalgorithm that executes entirely on the GPU, which allows to efficiently\nbatch-evaluate a whole population of networks. Within this framework, we\nexplore the limited evaluation evolutionary algorithm for neural network\ntraining and find that its batch evaluation idea comes with a large accuracy\ntrade-off. In further experiments, we explore crossover operators and find that\nunprincipled random uniform crossover performs extremely well. Finally, we\ntrain a network with 92k parameters on MNIST using an EA and achieve 97.6 %\ntest accuracy compared to 98 % test accuracy on the same network trained with\nAdam. Code is available at https://github.com/jprellberg/gpuea.","url_abs":"http://arxiv.org/abs/1806.09819v1","url_pdf":"http://arxiv.org/pdf/1806.09819v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"limited-evaluation-evolutionary-optimization","repo_url":"https://github.com/jprellberg/gpuea","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"evolutionary-algorithms","task_name":"Evolutionary Algorithms"},{"task_slug":null,"task_name":"GPU"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}