Federal gosudarstvennoe uchebnoe predpriyatie Chair of the System of Artificial Intelligence

Chapter 1
topology.
Application of technologies of neural networks for the solution of digital problems allows to achieve a high level
of parallelism and makes it possible to implement algorithms using distributed digital complexes, systems and digital
networks.
Neyronnye seti ÿ udobnÿy i dostatochno prostoy instrument dlya sozdaniya razlichnyx ekspertnyx sistem,
resheniya zadach klassifikatsii i izvlecheniya znaniy iz dannyx i, krome togo, dayut vozmojnost vysokoparallelnyx
realizatsiy. Rassmatrivaya neyronnuyu set kak nabor ÿlementov, proizvodyashchix nekotorÿe vÿchisleniya nad
prixodyashchimi k nim dannymi, mojno izuchat reshenie glavnoy i vspomogatelnyx zadach. K glavnoy otnosyatsya
funktsii generatsii, obucheniya i testirovaniya neyronnyx setey. Vspomogatelnÿe tasks: definition of significance of input
signals, contrast of neural networks, training of primeers, definition of minimum reshayushchego set of input parameters,
receipt of logical transverse neural network and, finally, schema. Dlya resheniya etix zadach razrabotan i ispolzuetsya
matematicheskiy apparat iskusstvennyx neyronnyx setey. Architecture and algorithms pozvolyayut obuchat i testirovat
neyronnye seti, reshat razlichnÿe zadachi obrabotki
Odnako, poskolku bolshinstvo polzovateley NKS poka ne mogut poluchit pryamogo dostupa k parallelnym
sistemam, postolku v dannoy rabote idet obsujdenie intellektualnyx vozmojnostey i preimushchestv neyronnyx setey. V
bolshinstve obsujdaemyx zadach i ix resheniy ispolzovany iskusstvennÿe neyronnÿe seti pryamogo rasprostraneniya,
po zarubejnym klassifikatsiyam otnosyashchiesya k klassu MLP (mnogosloynyy perseptron). Takje neyronnÿe setey
takogo vida nazÿvayut dvoystvennÿe neyronnye seti iz-za chasto ispolzuemyx metodov obucheniya neyronnyx setey
takogo vida, sigmoidnÿe neyronnye seti - iz-za preobladayushchego ispolzovaniya sigmoidnoy (sigmoidalnoy,
aktivovidhesti) funktsii Neyronnye seti predstavlyayut soboy sovokupnost odnotipnyx bazovyx ÿleterov - neyronov,
soedinennyx mejdu soboy liniyami peredachi informatsii ili sinapticheskimi svyazyami s opredelennymi koeffitsientami
vesa i, blagodarya takoy strukture, obladayushchigo step vlen. Vÿdelyayut gruppu svyazey, po kotorym set poluchaet
informatsiyu iz vneshnego mira, i gruppu vyxodnyx svyazey, s kotoryx snimayutsya vÿdavaemÿe setyu signalaly.
Neyronnye set primenyayutsya dlya resheniya razlichnyx zadach klassifikatsii i prognozirovaniya. Neyronnaya set
obuchaetsya resheniyu zadachi na osnovanii
10
Machine Translated by Google

---

neurocomputers.
mnogo, sushchestvuet mnogestvo modeley, imitiruyushchix vneshnie proyavleniya
element of the nervous system. Izuchenie mechanisms of function
obrabotki informatsii v raspredelennyx neyronnyx setyax, nazÿvayut
about 1015 peredayushchix svyazey, imeyushchix dlinu do metra i
bolee. Biological and artificial neurons. Nervous cell,
sokrashchenno nazÿvaemaya neuronom, yavlyaetsya osnovnÿm
"vxod – trebuemÿy vyxod", i dalee sposobna reshat primery, ne vxodyashchie v
Biological neurons and neural networks serve as prototypes
poznaniya protekayushchix v nervnoy sisteme protsessov poiska, peredachi i