Přístupy meta-učení pro automatické strojové učení
Název práce v češtině: | Přístupy meta-učení pro automatické strojové učení |
---|---|
Název v anglickém jazyce: | Meta-learning approaches for automated machine learning |
Klíčová slova: | strojové učení, meta-učení, auto-ML, evoluční algoritmy |
Klíčová slova anglicky: | machine learning, meta-learning. auto-ML, evolutionary computing |
Akademický rok vypsání: | 2020/2021 |
Typ práce: | disertační práce |
Jazyk práce: | čeština |
Ústav: | Ústav informatiky AV ČR, v.v.i. (32-UIAV) |
Vedoucí / školitel: | Mgr. Roman Neruda, CSc. |
Řešitel: | skrytý - zadáno a potvrzeno stud. odd. |
Datum přihlášení: | 08.09.2020 |
Datum zadání: | 08.09.2020 |
Datum potvrzení stud. oddělením: | 30.09.2020 |
Zásady pro vypracování |
Automated machine learning can be cast as the so-called Combined Algorithm Selection and Hyperparameter (CASH) optimization problem. To solve this problem, an efficient search procedure is needed to operate on the complex space of machine learning workflows descriptions together with hyper-parameter optimization. One of possible approaches can utilize meta-learning principles of learning from previous experience.
The goal of the work is to study suitable representations of machine learning workflows and to propose a general framework for their optimization based on evolutionary learning techniques. The representation should be flexible enough to include standard ML models and their combinations, such as ensembles, and it should also include possibility for efficient hyper-parameter search. Meta-learning approach will be extended for machine learning workflows in a way it enables to speed up the search procedure either by bootstrapping or constraining the optimization process. |
Seznam odborné literatury |
Hutter, Frank and Kotthoff, Lars and Vanschoren, Joaquin (Eds.): Automated Machine Learning: Methods, Systems, Challenges. Springer, 2019, (ISBN 978-3-030-05318-5).
Brazdil, P. and Giraud Carrier, C. and Soares, C. and Vilalta, R.: Metalearning - Applications to Data Mining. Springer, 2009, (ISBN 978-3-540-73263-1). Xin He and Kaiyong Zhao and Xiaowen Chu: AutoML: A Survey of the State-of-the-Art. arXiv:1908.00709 [cs.LG], 2019, (https://arxiv.org/abs/1908.00709). Zöller, Marc-André and Huber, Marc-André: Benchmark and Survey of Automated Machine Learning Frameworks. arXiv:1904.12054 [cs.LG], 2019, (https://arxiv.org/abs/1904.12054). |