Excess Entropies Suggest the Physiology of Neurons to Be Primed for Higher-Level Computation.

Autor:	Stoop RL; Institute of Neuroinformatics, University and ETH Zürich, Irchel Campus, Winterthurerstrasse 190, 8057 Zürich, Switzerland., Stoop N; Department of Mathematics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139-4307, USA.; Institute of Building Materials, ETH Zürich, Stefano-Franscini-Platz 3, 8093 Zürich, Switzerland., Kanders K; Institute of Building Materials, ETH Zürich, Stefano-Franscini-Platz 3, 8093 Zürich, Switzerland., Stoop R; Institute of Neuroinformatics, University and ETH Zürich, Irchel Campus, Winterthurerstrasse 190, 8057 Zürich, Switzerland.
Jazyk:	angličtina
Zdroj:	Physical review letters [Phys Rev Lett] 2021 Oct 01; Vol. 127 (14), pp. 148101.
DOI:	10.1103/PhysRevLett.127.148101
Abstrakt:	Biological neuronal networks excel over artificial ones in many ways, but the origin of this is still unknown. Our symbolic dynamics-based tool of excess entropies suggests that neuronal cultures naturally implement data structures of a higher level than what we expect from artificial neural networks, or from close-to-biology neural networks. This points to a new pathway for improving artificial neural networks towards a level demonstrated by biology.
Databáze:	MEDLINE
Externí odkaz:	Zobrazit plný text záznamu