{"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/deep-super-learner-a-deep-ensemble-for","title":"Deep Super Learner: A Deep Ensemble for Classification Problems","arxiv_id":"1803.02323","date":"2018-03-06","proceeding":null,"authors":["Steven Young","Tamer Abdou","Ayse Bener"],"abstract":"Deep learning has become very popular for tasks such as predictive modeling\nand pattern recognition in handling big data. Deep learning is a powerful\nmachine learning method that extracts lower level features and feeds them\nforward for the next layer to identify higher level features that improve\nperformance. However, deep neural networks have drawbacks, which include many\nhyper-parameters and infinite architectures, opaqueness into results, and\nrelatively slower convergence on smaller datasets. While traditional machine\nlearning algorithms can address these drawbacks, they are not typically capable\nof the performance levels achieved by deep neural networks. To improve\nperformance, ensemble methods are used to combine multiple base learners. Super\nlearning is an ensemble that finds the optimal combination of diverse learning\nalgorithms. This paper proposes deep super learning as an approach which\nachieves log loss and accuracy results competitive to deep neural networks\nwhile employing traditional machine learning algorithms in a hierarchical\nstructure. The deep super learner is flexible, adaptable, and easy to train\nwith good performance across different tasks using identical hyper-parameter\nvalues. Using traditional machine learning requires fewer hyper-parameters,\nallows transparency into results, and has relatively fast convergence on\nsmaller datasets. Experimental results show that the deep super learner has\nsuperior performance compared to the individual base learners, single-layer\nensembles, and in some cases deep neural networks. Performance of the deep\nsuper learner may further be improved with task-specific tuning.","url_abs":"http://arxiv.org/abs/1803.02323v1","url_pdf":"http://arxiv.org/pdf/1803.02323v1.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":"deep-super-learner-a-deep-ensemble-for","repo_url":"https://github.com/levyben/DeepSuperLearner","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}