Federal gosudarstvennoe uchebnoe predpriyatie Chair of the System of Artificial Intelligence

Control questions to 1 glave:
After obucheniya neyronnoy seti neobxodimo provesti ee testirovanie
On osnovanii obuchayushchey vyborki dostatochno slojno opredelit, skolko
sloev i neyronov seti neobxodimo. Poetomu postupayut obÿchno tak. Obuchayut set
so strukturoy, predlagaemoy programmoy-neyroimitatorom po
(odinakovymi vxodnymi vektorami dannyx, no raznymi vyxodnymi). Takim
neuroset. 3.
Obuchenie neyroseti. Ponyatie, naznachenie, funktsionirovanie
set larger size. On the practice of reshenii raznoobraznyx zadach
razresheniya takoy protivorechivoy situatsii. Poyavlenie takix konfliktnyx
obuchayushchuyu vÿborku zadach. Tochnost pravilnogo resheniya ochen silno
priznakov, poskolku pri rasshirenii priznakovogo prostranstva
Consider the factors, from which depend on the success of training
sotney neyronov - obychno xvataet neskolkix desyatkov neyronov i daje
razlichnyx neformalizovannyx zadach v raznyx problemnyx oblastyax
doljna byt dostatochno gibkoy, chtoby nauchitsya pravilno reshat vse
Odnako, daje uvelichenie razmera neyronnoy seti ne pomojet, esli
dobavlyaemogo priznaka i kriticheskaya situatsiya budet ischerpana. In love
protsentu pravilnyx otvetov pri reshenii etix je zadach spetsialistom ekspertom.
isklyucheniem konfliktnyx primerov iz zadachnika.
dostatochnoe kolichestvo neyronov i svyazey.
prisutstvuyut zadachi s odinakovymi usloviyami, no raznymi otvetami
1. Neuron. Ponyatie, naznachenie, funktsionirovanie. Types of neurons. 2.
Neuroset. Ponyatie, naznachenie, funktsionirovanie. Tipy
umolchaniyu, a v dalneyshem, esli set ne mojet obuchitsya, probuyut obuchit
zadacham neyronnaya set obuchitsya ne mojet. Here voznikaet problem
na testovoy vÿborke dlya opredeleniya tochnosti resheniya ne vxodivshix v
prakticheski ne vstrechaetsya situatsii, kogda trebuetsya neyroset s bolee chem
1.6.3 Factors influencing the learning of the neural network
primerov mojet, dopustim, oznachat nedostatochnost nabora vxodnyx
zavisit ot reprezentativnosti obuchayushchey vÿborki. Obychno pri reshenii
49
Machine Translated by Google

---

dostoinstva i nedostatki. 9.
Elements of the neuroconstructor
4. Gradient and sluchaynye method. Sxodstvo and razlichie.
5. Hardware and software realization of neurons. 6. Ukazat
tip zadach, reshaemyx s pomoshchyu neyrosetey. 7. Basic
parameters, influencing the effectiveness of training and
10. Contrast: ponyatie, naznachenie.
work neuroset. 8.
Obuchenie po zadachniku i po otdelnomu primeru: sravnit
50
Machine Translated by Google

---

experiments. Osnovnÿe tipy zadach, reshaemyx s pomoshchyu neyronnyx
nakopleno dostatochnoe number of
primerov; the problem is characterized by large volumes of input information;
dannÿe nepolnÿ ili izbytochnÿ, zashumlenÿ, chastichno protivorechivy.
Neyronnye seti dlya polzovatelya PK - instrument, pozvolyayushchiy
Vÿpolnenie issledovaniya s pomoshchyu neyronnyx setey mojno
mogut byt ispolzovany dlya obrabotki rezultatov razlichnyx
2.1.1. Pre-processing operations