Federal gosudarstvennoe uchebnoe predpriyatie Chair of the System of Artificial Intelligence

1.00 4
0.14 27
X10 0.43 4
0.10 8
0.15 24
Set
0.05
3
5
5
1
0.14 8
0.28 3
0.16 5
8
(znachimosti usrednyayutsya vo vsem setyam), tem ne menee rang pokazÿvaet
0.15
Sre
1
0.15 25
ST 0.03 10 0.18 5
0.27 4
0.22 6
Rang
6
0.36 2
0.38 10
2
Set
0.15 25
0.16 6
5
7
12
dnee
3
menee 50% correctly identified primerov, t. k. v etom sluchae vliyanie
8
0.28 3
bolshuyu znachimost dannogo parameter. Rang tochnee opredelyaet poryadok
A 0.53 3
7
0.38 2
1
Rang
6
2
0.23
4
5
0.18 3
4
9
parameters weak or practically otsutstvuet. There is no menee vozmojno
0.16 5
LIT 0.03 10 0.18 5
0.21 3
sledovaniya parametrov po stepeni ix vliyaniya na vyxodnoe pole.
1
5
2
7
0.22 6
5
ma
SL 0.04 9
Table 3.3. - Results opredeleniya znachimostey vxodnyx
0.29 14
Set
SV 0.03 10 0.19 4
0.15 25
0.23
AU
ran
Rang
10
konstruirovanie bolee tonkix nastroek seti.
Rekomenduetsya dopolnitelno (pomimo znacheniy) opredelyat color
5
0.10 8
0.09 9
0.18 5
parameters
3
1.00 1
3
SU 0.09 5
0.14 6
T
gov
5
vxodnyx parameters for kajdoy seti. It pozvolyaet bolee adequately
0.17 4 0.15 24
11
X9 1.00 1
0.14 8
Set
0.15 25
2
4
0.36 2
1.00 1
1.00 1
Rang
0.15 26
E
0.16 5
Sum
provodit analiz vliyaniya parametrov na vyxodnoe pole. In bolshinstve
0.23
4
WIN 0.08 6
4
0.07 7
0.22 6
Rang
102
Machine Translated by Google

---

0.03 12 0.03 45
0.03 44
0.01 11 0.02 13 0.02 49 20
okrugliv znacheniya do sotyx. Rasstavlyaem rang dlya kajdogo znacheniya,
nachinaya s maksimalnogo: parametru, imeyushchemu naibolshee znachenie
0.05 10 0.04 38
0.13
19
MN 0.04 9
0.03 9
FR 0.04 9
0.03 9
okrugleniya) imeyut odinakovÿe rangi. Prostaviv takim obrazom rang dlya
0.21 7
0.03 9
0.04 11
15
0.03 12 0.02 10 0.04 11 0.04 39
0.03 41
X15 0.03 10 0.02 13
12
0.02 13
11
8
ubyvaniyu (ot bolshix k menshim) for kajdoy seti, predvaritelno
ZAR 0.04 12 0.04 11
0.03 9
0.11 7
ST 0.03 10 0.05 10 0.04 8
SPR 0.07 7
X16 0.03 10 0.04 11
13
X17 0.02 11 0.02 13 0.02 10 0.02 13 0.02 47
0.04 11
KR 0.02 11 0.18 5
8
0.04 11
X12 0.01 12 0.02 13
9
X27 0.02 11 0.02 13
0.05 10 0.04 40
14
znachimomu - 2 i t.p. Odinakovÿe znacheniya znachimostey (with uchetom
neyrosetey: znacheniya znachimostey vxodnyx parametrov uporyadochivaem po
0.27 4
0.04 11
X30 0.08 6
0.03 43
19
18
0.04 11 0.04 39
0.14 29
X28 0.02 11 0.02 13
13
19
TH 0.04 9
0.03 9
14
9
0.05 10 0.04 42
0.03 12 0.03 9
0.04 11 0.04 41
17
znachimosti (in NeuroPro eto obÿchno 1) stavim v sootvetstvie rang 1, menee
WN 0.05
7
0.03 12 0.03 9
9
10
0.02 13
X29 0.01 12 0.03 12 0.02 10 0.02 13 0.02 47
0.03 12 0.03
0.14 6
16
0.03 9
0.03 12 0.03 41
Predlagaetsya sleduyushchaya metodika dlya opredeleniya rangov po serii
X14 0.03 10 0.03 12 0.04 8
0.18 36
0.04 11 0.02 10 0.03 12 0.03 42
0.02 13 0.02 46
11
0.15 27
X2 0.02 11 0.02 13
13
0.01 11 0.02 13 0.02 47
0.03 9
AV 0.66 2
TU 0.04 9
103
Machine Translated by Google

---

summarnogo ranga (from minimalnogo znacheniya summarnogo ranga do
kontrastirovaniyu (sokrashcheniyu) vnutrenney libo vneshney struktury seti. At this
posledovatelno udalyaetsya odin element seti (naprimer, synaps, neuron, neodnorodnyy
vxod pri kontrastirovanii vnutrenney struktury) or vxodnoy signal (sokrashchenie chisla
vxodnyx parametrov), zatem set
sootvetstvuyushchiy item menu) vÿpolnyayutsya neyroimitatorom
sootvetstvii s vozrastaniem summarnogo ranga: minimalnaya summa rangov
The process of sokrashcheniya number of elements or inputs is repeated, if it is set
Dlya etogo sleduet:
1. Otkrÿt proekt, poluchennÿy pri obuchenii po odnoy iz vÿborok
n_ok2).
3. Open the selection.
4. Provesti kontrastirovanie dlya kajdoy neyroseti. 5. Save
the quantity and percentage (of the total quantity) udalennyx
tablitsy dlya opredeleniya znachimostey po 5 setyam - tabl. 3.3)
ne udalyaetsya. After the end of the process of contrasting the structure of the network
parameters, bolee adequately used when analyzing vliyaniya parameters
signals remain unchanged. At further training (on drugoy
perekrestnoe testirovanie analogichno privedennoy metodike s soxraneniem
kontrastirovaniya posle nachala protsessa (user vÿbral
setey, kotoryy rasschityvaetsya kak summa rangov kajdoy seti. Uporyadochivaem
parameters in sootvetstvii with poryadkom vozrastaniya
tochnee, chem po usrednennym znacheniyam znachimostey).
G2. Uproshchenie seti. Mojno otdelno provesti eksperimenty po
contrast procedures. At the definition of znachimostey and rangov posle
maximum). Itogovyy rang po serii neyrosetey prostavlyaetsya v
douchivaetsya. Esli pri etom kachestvo obucheniya sootvetstvuet trebuemomu, to
avtomaticheski.
(obshchey, vyborok 1 or 2).
2. Save the project under the drug name (for example, n_ok, n_ok1 or
sootvetstvuet rank 1, chut bolshe minimalnogo - 2 i t.p. (primitive type
ne v sostoyanii douchitsya, to protsess uproshcheniya ostanavlivaetsya i element
Sleduet otmetit, chto rangi, v otlichie ot srednix znachimostey
(pri uproshchenii vnutrenney struktury) i chislo ostavshixsya vxodnyx
elements in vide tablitsy. 6.
Save the project under tekushchim imenem.
After kontrastirovaniya mojno provesti testirovanie ili
(on the extreme mere, poryadok sledovaniya znachimyx parameters po rangam
kajdoy seti, for kajdogo parameter opredelyaem summarnyy rang dlya serii
vÿborke dannyx) ispolzuetsya sokrashchennÿy set of parameters. Operations
results (Table 3.4).
Determining the characteristics of the input parameters can be done later
104
Machine Translated by Google

---

36.64 1.79
Set 8
174 66.41
64.12
primerov
analog method (see G1). No for parameters, vÿrezannyx pri
145
77.78
33.59 1.78
Set 6
35.88
9.54
Table 3.4. - Results perekrestnogo testirovaniya posle
1.76
28
46.67
ka
30
58.40
82,22
Max.
increase
30,15
vxodov
168
62,22
96
vremennyx resursov rekomenduetsya srazu vÿrezat iz tablitsy so
reshennyx
kontrastirovaniya vneshney struktury (vxodnyx parameters) ispolzuetsya
88
35
6.16
69.85
35.11
vid tablitsy dlya 5 setey posle kontrastirovaniya - tabl. 3.5).
Set 1
1.73
66.79
Sred
94
37
1.70
9.40
166 63.36
maksimalnÿy rang, uvelichennyy na 1. For effective use
Set 4
increase
7.87
21
153
183
5.70
79
Udaleno
71.11
28
175
92
30.53
66.67
primerov
33.21
Set 2
Set 3
Srednee 30.6 68.00
5.87
kol-vo% kol-vo% kol-vo%
nulevoe srednee znachenie pokazateley znachimosti po vsem setyam (primernyy
109 41.60 2.03
kontrastirovanii, znachimosti ne vÿchislyayutsya (v tablitse znachenie 0). Takim
parameters stavitsya in sootvetstvie, vÿbrannÿy iz rangov vsex setey
nyaya
Set 7
55.34
7.02
Neverno
contrast
32
62,22
Set 5
182 69.47
117 44.66 2.14
reshennyx
6.68
30
1.84
ka
znachimostyami posle kontrastirovaniya vxodnÿe parameters, imeyushchie
80
66.67
170 64.89
9.48
87
Verno
1.67
105
Machine Translated by Google

---

Rang
7
45
7
17