{ "article_id": 1457, "public_id": "7bc430f2-e965-4992-9244-0012ec7a4405", "internal_identifier": null, "source_identifier": "90:6", "status": "published", "type": "research_article", "title": "İleri Beslemeli Yapay Sinir Ağları için Donanım Uygulamasına Yatkın Çevrimiçi ELM Tabanlı Eğitim Algoritması", "title_secondary": null, "abstract": "Bu bildiride, ileri beslemeli tek gizli katmanlı yapay sinir ağları için donanım uygulamasına uyumlu olan ve seyrek düzenlileştirilmiş bir aşırı öğrenme makinası (ELM) algoritması sunulmaktadır. Önerilen algoritma, düzenlileştirme için genelleştirilmiş minimaks konkav (GMC) ceza fonksiyonuna dayanmakta ve matris tersi alma veya norm hesaplaması gibi donanım mimarisi için karmaşıklığı arttıracak matematiksel işlemleri içermemektedir. Önerilen algoritma, çevrimiçi sıralı ELM (OS-ELM) algoritmasına benzer olarak, çevrimiçi öğrenmeyi destekleyecek şekilde değiştirilmiştir. Önerilen algoritma, UCI makine öğrenmesi veri deposundan alınan veri kümeleri ile test edilmiş ve genel olarak daha yüksek seyreklik seviyelerine ulaşmıştır. OS-ELM algoritmasına göre genellikle daha iyi sonuçlar vermiştir. Ek olarak, düzenlileştirme ile artan seyreklik seviyeleri sayesinde, önerilen algoritma yapay sinir ağından gereksiz değişkenleri elemekte ve test verileri için tahmin doğruluğunu artırmaktadır.", "language": "tr", "english_title": "Hardware-Application-Compatible Online ELM-Based Training Algorithm for Feedforward Artificial Neural Networks", "english_abstract": "This paper presents a sparsely regularized extreme learning machine (ELM) algorithm compatible with hardware implementation for feedforward single-hidden-layer artificial neural networks. The proposed algorithm relies on a generalized minimax concave (GMC) penalty function for regularization and does not include mathematical operations that would increase complexity for the hardware architecture, such as matrix inversion or norm calculation. Similar to the online sequential ELM (OS-ELM) algorithm, the proposed algorithm has been modified to support online learning. The proposed algorithm was tested with datasets from the UCI machine learning data repository and generally achieved higher sparsity levels, yielding generally better results than the OS-ELM algorithm. Additionally, thanks to the increased sparsity levels through regularization, the proposed algorithm eliminates redundant variables from the artificial neural network and improves prediction accuracy for the test data.", "doi": "10.36287/setsci.6.1.013", "doi_url": "https://doi.org/10.36287/setsci.6.1.013", "resource_url": "https://www.set-science.com/a/7bc430f2-e965-4992-9244-0012ec7a4405", "pdf_url": "https://www.set-science.com/a/7bc430f2-e965-4992-9244-0012ec7a4405/pdf", "og_image": "https://www.set-science.com/assets/img/og-card-v2.png", "authors": [ { "public_id": "6a7498b8-2aa5-4331-90dd-cba1df2e6eeb", "given": "Önder", "family": "Polat", "full": "Önder Polat", "literal": null, "orcid": null, "orcid_url": null, "email": "onderp@gantep.edu.tr", "corresponding": false, "sequence": "first", "affiliations": [ { "name": "Gaziantep Üniversitesi", "department": null, "country": null, "country_code": null, "ror": null, "ror_url": null } ] }, { "public_id": "cf4b54df-3597-45fb-9949-9a8fc1989663", "given": "Sema", "family": "Koç Kayhan", "full": "Sema Koç Kayhan", "literal": null, "orcid": null, "orcid_url": null, "email": "skoc@gantep.edu.tr", "corresponding": true, "sequence": "additional", "affiliations": [ { "name": "Gaziantep Üniversitesi", "department": null, "country": null, "country_code": null, "ror": null, "ror_url": null } ] } ], "keywords": [ "Yapay Sinir Ağı", "Çevrimiçi makine öğrenmesi", "ELM", "Donanım uygulaması" ], "translations": [ { "language": "en", "title": "Hardware-Application-Compatible Online ELM-Based Training Algorithm for Feedforward Artificial Neural Networks", "abstract": "This paper presents a sparsely regularized extreme learning machine (ELM) algorithm compatible with hardware implementation for feedforward single-hidden-layer artificial neural networks. The proposed algorithm relies on a generalized minimax concave (GMC) penalty function for regularization and does not include mathematical operations that would increase complexity for the hardware architecture, such as matrix inversion or norm calculation. Similar to the online sequential ELM (OS-ELM) algorithm, the proposed algorithm has been modified to support online learning. The proposed algorithm was tested with datasets from the UCI machine learning data repository and generally achieved higher sparsity levels, yielding generally better results than the OS-ELM algorithm. Additionally, thanks to the increased sparsity levels through regularization, the proposed algorithm eliminates redundant variables from the artificial neural network and improves prediction accuracy for the test data.", "keywords": [ "Artificial Neural Network", "Online Machine Learning", "ELM", "Hardware Application" ] } ], "subjects": [], "subject_names": [], "published_date": "2023-12-29", "published_year": "2023", "received_date": null, "accepted_date": null, "volume": "15", "issue": null, "page_first": "27", "page_last": "32", "article_number": null, "conference": { "short_name": "BMYZ2023", "full_title": "Cognitive Models and Artificial Intelligence Conference", "name": "Cognitive Models and Artificial Intelligence Conference", "acronym": "BMYZ2023", "location": "Ankara, Türkiye", "start_date": "2023-10-26", "end_date": null, "public_id": "dc3a47c3-f4b3-4059-b610-f99aca33d7a0" }, "proceedings": { "title": "Cognitive Models and Artificial Intelligence Conference", "series_title": "SETSCI Conference Proceedings", "publisher": "SET Technology", "isbn": "978-605-69925-7-5", "eisbn": null, "issn": null, "eissn": "2687-5527", "doi": null, "doi_prefix": "10.36287", "slug": "setsci-volume-15-2023-legacy-90", "public_id": "33fc699d-0ab3-4057-b2e0-644dc89f900d", "url": "https://www.set-science.com/p/setsci-volume-15-2023-legacy-90" }, "license": { "name": "CC BY 4.0", "url": "https://creativecommons.org/licenses/by/4.0/", "code": "CC-BY-4.0", "open_access": true }, "copyright": { "holder": "SET Technology", "year": "2023" }, "funding_statement": null, "funders": [], "references": [ { "order": 1, "raw": "G. B. Huang, H. Zhou, X. Ding, ve R. Zhang, “Extreme Learning Machine for Regression and Multiclass Classification”, IEEE Trans. Syst. Man Cybern. Part B Cybern., c. 42, sy 2, ss. 513-529, Nis. 2012, doi: 10.1109/TSMCB.2011.2168604.\r\n\r\nG.-B. Huang, D. H. Wang, ve Y. Lan, “Extreme learning machines: a survey”, Int. J. Mach. Learn. Cybern., c. 2, sy 2, ss. 107-122, Haz. 2011, doi: 10.1007/s13042-011-0019-y.\r\n\r\nG.-B. Huang, Q.-Y. Zhu, ve C.-K. Siew, “Extreme learning machine: Theory and applications”, Neurocomputing, c. 70, sy 1, ss. 489-501, Ara. 2006, doi: 10.1016/j.neucom.2005.12.126.\r\n\r\nYoan Miche, A. Sorjamaa, P. Bas, O. Simula, C. Jutten, ve A. Lendasse, “OP-ELM: Optimally Pruned Extreme Learning Machine”, IEEE Trans. Neural Netw., c. 21, sy 1, ss. 158-162, Oca. 2010, doi: 10.1109/TNN.2009.2036259.\r\n\r\nB. Efron, T. Hastie, I. Johnstone, ve R. Tibshirani, “Least angle regression”, Ann. Stat., c. 32, sy 2, ss. 407-499, Nis. 2004, doi: 10.1214/009053604000000067.\r\n\r\nH.-J. Rong, Y.-S. Ong, A.-H. Tan, ve Z. Zhu, “A fast pruned-extreme learning machine for classification problem”, Neurocomputing, c. 72, sy 1-3, ss. 359-366, Ara. 2008, doi: 10.1016/j.neucom.2008.01.005.\r\n\r\nZ. Shao, M. J. Er, ve N. Wang, “An Efficient Leave-One-Out Cross-Validation-Based Extreme Learning Machine (ELOO-ELM) With Minimal User Intervention”, IEEE Trans. Cybern., c. 46, sy 8, ss. 1939-1951, Ağu. 2016, doi: 10.1109/TCYB.2015.2458177.\r\n\r\nY. Miche, M. van Heeswijk, P. Bas, O. Simula, ve A. Lendasse, “TROP-ELM: A double-regularized ELM using LARS and Tikhonov regularization”, Neurocomputing, c. 74, sy 16, ss. 2413-2421, Eyl. 2011, doi: 10.1016/j.neucom.2010.12.042.\r\n\r\nÖ. Polat ve S. K. Kayhan, “GPU-accelerated and mixed norm regularized online extreme learning machine”, Concurr. Comput. Pract. Exp., c. 34, sy 15, s. e6967, 2022, doi: 10.1002/cpe.6967.\r\n\r\nH. T. Huynh ve Y. Won, “Regularized online sequential learning algorithm for single-hidden layer feedforward neural networks”, Pattern Recognit. Lett., c. 32, sy 14, ss. 1930-1935, Eki. 2011, doi: 10.1016/j.patrec.2011.07.016.\r\n\r\nZ. Shao ve M. J. Er, “An online sequential learning algorithm for regularized Extreme Learning Machine”, Neurocomputing, c. 173, ss. 778-788, Oca. 2016, doi: 10.1016/j.neucom.2015.08.029.\r\n\r\nT. Song, D. Li, Z. Liu, ve W. Yang, “Online ADMM-Based Extreme Learning Machine for Sparse Supervised Learning”, IEEE Access, c. 7, ss. 64533-64544, 2019, doi: 10.1109/ACCESS.2019.2915970.\r\n\r\nC.-H. Zhang ve J. Huang, “The sparsity and bias of the Lasso selection in high-dimensional linear regression”, Ann. Stat., c. 36, sy 4, ss. 1567-1594, Ağu. 2008, doi: 10.1214/07-AOS520.\r\n\r\n“A Survey on Nonconvex Regularization-Based Sparse and Low-Rank Recovery in Signal Processing, Statistics, and Machine Learning | IEEE Journals & Magazine | IEEE Xplore”. Erişim: 11 Ekim 2023. [Çevrimiçi]. Erişim adresi: https://ieeexplore.ieee.org/document/8531588\r\n\r\nC.-H. Zhang, “Nearly unbiased variable selection under minimax concave penalty”, Ann. Stat., c. 38, sy 2, ss. 894-942, Nis. 2010, doi: 10.1214/09-AOS729.\r\n\r\nI. Selesnick, “Sparse Regularization via Convex Analysis”, IEEE Trans. Signal Process., c. 65, sy 17, ss. 4481-4494, Eyl. 2017, doi: 10.1109/TSP.2017.2711501.\r\n\r\nH. H. Bauschke ve P. L. Combettes, Convex Analysis and Monotone Operator Theory in Hilbert Spaces. içinde CMS Books in Mathematics. New York, NY: Springer, 2011. doi: 10.1007/978-1-4419-9467-7.\r\n\r\n“Citation - UCI Machine Learning Repository”. Erişim: 11 Ekim 2023. [Çevrimiçi]. Erişim adresi: https://archive-beta.ics.uci.edu/citation", "doi": null, "title": null, "authors": null, "source": null, "year": null, "volume": null, "pages": null } ], "statements": { "acknowledgements": null, "conflict_of_interest": null, "data_availability": null, "ai_disclosure": null }, "credit": [], "history": { "published": "2023-12-29" } }