Optimalizace architektur neuronových sítí pomocí evolučních algoritmů

• Thesis title in Czech: Optimalizace architektur neuronových sítí pomocí evolučních algoritmů
• Thesis title in English: Neural architecture search by means of evolutionary computing
• Key words: Hluboké učení, hledání neuronových architektur, automatické strojové učení, evoluční algoritmy.
• English key words: Deep learning, neural architecture search, auto-ML, evolutionary computing.
• Academic year of topic announcement: 2020/2021
• Thesis type: dissertation
• Thesis language: čeština
• Department: Ústav informatiky AV ČR, v.v.i. (32-UIAV)
• Supervisor: Mgr. Roman Neruda, CSc.
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 08.09.2020
• Date of assignment: 08.09.2020
• Confirmed by Study dept. on: 06.10.2020

Guidelines

Complex and specialized architectures of deep neural networks that have been successful in solving practical tasks motivated a recent research in a subfield of automated machine learning called neural architecture search (NAS). State of the art approaches to this problem include reinforcement learning, Bayesian optimization, and evolutionary computing. The goal of this work is to study efficient block representations of current deep neural networks, and to develop an efficient search procedure based on evolutionary computing. The student should explore direct acyclic graph structure of modules in deep networks suitable for optimization approaches such as tree-based or cartesian genetic programming. Recent approaches to deep architectures encoding, such as hierarchical representations, parameter sharing, or differentiable search should be taken into account.

References

Hutter, Frank and Kotthoff, Lars and Vanschoren, Joaquin (Eds): Automated Machine Learning: Methods, Systems, Challenges, Springer, 2019.

Thomas Elsken and Jan Hendrik Metzen and Frank Hutter: Neural Architecture Search: A Survey. Journal of Machine Learning Research, 20 (55), 1-21, 2019.

Xin He and Kaiyong Zhao and Xiaowen Chu : AutoML: A Survey of the State-of-the-Art. arXiv:1908.00709 [cs.LG], 2019.

E. Real, S. Moore, A. Selle, S. Saxena, Y. L. Suematsu,J. Tan, Q. V. Le, and A. Kurakin. Large-scale evolution of image classifiers. In ICML, 2902–2911, 2017.

Risto Miikkulainen, Jason Liang, Elliot Meyerson, Aditya Rawal, Dan Fink, Olivier Francon, Bala Raju, Hormoz Shahrzad, Arshak Navruzyan, Nigel Duffy, Babak Hodjat: Evolving Deep Neural Networks. arXiv:1703.00548 [cs.NE], 2019.