Federal gosudarstvennoe uchebnoe predpriyatie Chair of the System of Artificial Intelligence

1.5 Neurocomputer
Odnonapravlennÿe mnogosloynye seti sigmoidnogo tipa
8) predobrabotchik obuchayushchey vyborki (normirovshchik!) 9)
obuchayushchaya vyborka
17) contrast or modulus of contrast NS
neurocomputer: 1)
osnovnoy
raspolojennÿe na odnom urovne neyronÿ funktsioniruyut nezavisimo drug
2) chef
12) teacher constructor
tipa poluchili naibolshee rasprostranenie na praktike v silu
otnositelnoy prostoty ix matematicheskogo opisaniya. Sobstvenno
istoriya INS nachalas s odnosloynyx setey dannogo tipa. Odnako, vozmojnosti
odnosloynyx setey krayne skromny, poskolku
rabotÿ sleduet neobxodimost sushchestvovaniya sleduyushchix blokov
11) otsenka, otsenshchik, module otsenki
16) constructor of learning algorithms
predlojeny effektyvnye algorithms ix obucheniya.
4) constructor NN 5)
neuroset
13) tutor (teacher)
14) interpreter answers ns
3) setting
32
Machine Translated by Google

---

Scheme NKS
Rice. 11. Dvusloynyy perceptron
33
Machine Translated by Google

---

vÿchislitelnoy matematike: reshenie sistem lineynyx uravneniy, reshenie
dopolnitelnyx usloviy. In kachestve takix usloviy mogut ispolzovatsya
iz samyx nujnyx dlya polzovateley zadach: zapolnenie propuskov v
«Xoroshie» i lyubaya iz nix podoydet v kachestve resheniya zadachi. Pri
objects neizvestny znacheniya nekotoryx priznakov. Neobxodimo kak-to ix
peremennyx) on konechnomu naboru znacheniy nevozmojno bez spetsialnyx
predpolagaetsya, chto tipichnÿe neyrosetevÿe realizatsii dostatochno
Mnojestvo zadach, reshaemyx s pomoshchyu neyronnyx setey, postoyanno
summÿ kvadratov vtoryx proizvodnyx - trebovanie maksimalnoy
minimiziruyut chislo sloev, kolichestvo neyronov i / ili chislo svyazey, a
trebovaniya minimizatsii razlichnyx funktsionalov, naprimer integrala