Skoči na glavni sadržaj

Stručni rad

https://doi.org/10.19279/TVZ.PD.2021-9-2-06

KNOWLEDGE AND LEARNING OF A SINGLE NEURON

Predrag Valožić ; Tehničko veleučilište u Zagrebu, Zagreb, Hrvatska, profesor u mirovini


Puni tekst: hrvatski pdf 927 Kb

str. 120-129

preuzimanja: 174

citiraj


Sažetak

The learning outcomes of a linear artificial neuron with different prior knowledge were analyzed. Based on a common mathematical model of neurons and the same template for learning, the properties of different learning models more or less mathematized are illustrated with examples. It has been shown that "more advanced" models are faster but more sensitive and rigid in the solutions offered. "Childish", basic learning models are slower but more universal and "imaginative". A common criterion for assessing the "learning outcome" is the correctness of the solution to the problem – the design of a recursive sine wave generator of certain features: amplitude and frequency.

Ključne riječi

Linear neuron; learning; linear combination; inverse; pseudoinverse; feedback

Hrčak ID:

273767

URI

https://hrcak.srce.hr/273767

Datum izdavanja:

20.7.2021.

Podaci na drugim jezicima: hrvatski

Posjeta: 918 *