Robust Optimal-Size Implementation of Finite State Automata with Synfire Ring-Based Neural Networks

0503688 - ÚI 2020 RIV CH eng C - Conference Paper (international conference)
Cabessa, Jérémie - Šíma, Jiří
Robust Optimal-Size Implementation of Finite State Automata with Synfire Ring-Based Neural Networks.
Artificial Neural Networks and Machine Learning – ICANN 2019: Theoretical Neural Computation. Proceedings, Part I. Cham: Springer, 2019 - (Tetko, I.; Kůrková, V.; Karpov, P.; Theis, F.), s. 806-818. Lecture Notes in Computer Science, 11727. ISBN 978-3-030-30486-7. ISSN 0302-9743.
[ICANN 2019. International Conference on Artificial Neural Networks /28./. Munich (DE), 17.09.2019-19.09.2019]
R&D Projects: GA ČR(CZ) GA19-05704S
Institutional support: RVO:67985807
Keywords : Recurrent neural networks * Threshold circuits * Finite state automata * Synfire rings
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Synfire rings are important neural circuits capable of conveying synchronous, temporally precise and self-sustained activities in a robust manner. We describe an optimal-size implementation of finite state automata with neural networks composed of synfire rings. More precisely, given any finite automaton, we build a corresponding neural network partly composed of synfire rings capable of simulating it. The synfire ring activities encode the successive states of the automaton throughout its computation. The robustness of the network results from its architecture, which is composed of synfire rings and duplicated core components. In addition, the network's size is asymptotically optimal: for an automaton with n states, the network has theta (√n) cells.
Permanent Link: http://hdl.handle.net/11104/0295498