|
3.3 Features of working technology with SNN
121
Machine Translated by Google
Requests for data retrieval in the internal format of statistics,
4. Zapuskaem mastera resheniya zadach. Sposob resheniya zadachi - basic
soxranyaem.
3. It is divided into three selections: testovuyu (in the window "Data set editor" these
lines are shown in blue color and are not used in training, neobxodimy for nezavisimogo
testirovaniya), dannye dlya proverki (Lines
Request about translation of data from text into numeric format., Najimaem
vÿdelenÿ krasnÿm, neobxodimÿe dlya korrektirovki oshibki v protsesse
«Net»
obucheniya) i obuchayushchuyu vyborku (vÿdelenÿ chernÿm).
122
Machine Translated by Google
5.Ukazÿvaem kakie polya budut vyxodnymi. Ukazÿvaem field "Answer".
6.Ukazÿvaem kakie polya budut vxodnymi
123
Machine Translated by Google
9.Parameters output results
7.Vybiraem prodoljitelnost process obucheniya - srednyaya.
8.Parametry soxraneniya set - soxranyat 10 setey.
124
Machine Translated by Google
1
setey, kol-ve vxodnyx signalov, maksimalnoy oshibke i td.
Linear
OTVET
0.108249
1876
1880
10 * MLP
01 RBF
2
1
T. OTVET E. OTVET 2
-
7
1860
03 MLP 04
Inputs
0.006701
1872
0.03498
0.0848
Linear
1
10
10
3
6
7
6
8
9
12
11
0.008776
1868
11.Itak, protestiruem 10-uyu set v okne Run data Set:
10.V okne network Set Editor see information about tipax poluchennyx
Linear
08
0.993299
1
7
Type
7
09 MLP
1.991224
0.006701
2.034984
1.0848
Error
0.008776
02 RBF
Error
0.5316963
0.4725896
0.3540201
0.286078
0.2569745
0.2457205
0.2377385
0.2348149
0.09045
0.06388
2
-0.01636
0.03498
0.0848
0.108249
Linear
05
Hidden Performance 1
1.009487 1 0.8559816
4 0.7215794 -
0.4827206 - 0.4302589
1 0.4319629 -
0.4368305 - 0.4301544
7 0.1840289 4 0.09797
-
0.983644
06 MLP 07
Najimaem «Finish».
illustration 10-month set:
1864
1.10825
0.01636
125
Machine Translated by Google
7
0.004923
0.005192
0.00259
1976
1980
2
2
1.999075
0.007878
0.006366
0.002664
0.003802
0.008841
0.002335
0.002664
-
1
-
4
9
-
2
1
1
1
1924
1.00039
7
2.002793
1.992122
7
1
2
1
1928
1932
1936
1940
1
2
1
-
1884
-
2
6
0.997717
-
0.018653
9
0.005987
7
1968
1972
1892
7
0.003802
0.008841
0.002335
2
1
2
6.41e-05
0.001784
1.999936
1.998216
0.000924
1.907946
1.736035
1.001185
1.00043
0.006366
1944
-
1964
0.005472
-6.41e-05
1992
1.018653
1.988685
1.000262
0.005987
1.004923
2.005192
1.00259
1.993736
1
0.000389
0.0161
0.002793
0.007878
1.993634
0.01131
0.000261
1
2 1
-
0.004923
0.005192
0.00259
0.006264
0.0161
0.005472
0.0009246
0.09205
0.736035
0.001185
0.000429
2
1896
1900
1904
1908
-
0.000389
0.002282
0.002793
4
1888
1912
1916
1
1948
1952
1956
1960
2.016099
0.994527
2
1920
1
0.006264
8
6
-0.09205
0.736035
0.001185
0.000429
0.994013
1
0.002282
2
0.001784
0.018653
-0.01131
0.000261
1.003802
2.008841
1.002335
0.997335
126
Machine Translated by Google
Te. OTVET
1.5 0.7071068
0.3219904
0.5855474
0.4140445
0.8280891 1
ÿÿÿÿ ÿÿÿÿÿ ÿÿÿÿÿÿÿÿÿÿÿ ÿÿÿÿÿÿ ÿÿÿ ÿÿÿÿÿÿÿÿ ÿÿÿÿÿÿÿÿÿÿ.
Regression statistics
Data Mean
Data SD
Error Mean
Error SD
=> Training error graph =>
Tr. OTVET
1.44
0.5066228
7.76e-05
0.007204
0.005322
0.01422
0.9999
Abs E. Mean
SD Ratio
Correlation
12. The process of obucheniya seti v sisteme STATISTICA Neural Networks is
accompanied by automatic pokazom tekushchey oshibki obucheniya i
And. OTVET
1.4 0.5477226
0.04215
0.05366
0.0490628
0.09797
0.9957543
vÿchislyaemoy nezavisimo ot nee oshibkoy na kontrolnom mnojestve, pri
Run single case =>
127
Machine Translated by Google
distribution SNN. 1)
Open the database of Credit.sta
3.3.2 Primer 2
dannyx Credit.sta This file dannyx has a text format, attached k
neskolko ekranov pri ispolzovanii neyronnyx setey s otvetami, file
medlennee.
luchshe spravlyayutsya s zadaniem, although po sravneniyu s drugimi rabotayut
obÿchno, nachinaetsya s ispolzovaniya file dannyx. Nije pokazany
razlichnÿe sposoby testirovaniya dannyx, Zametim, chto seti MLP namnogo
Nachalo raboty s neyroimitatorom Statistica Neural Networks, kak
Output:
In the process of working with STATISTICA Neural Networks was used
128
Machine Translated by Google
3) Select the setting: Type solution -> "Standard"
vybor nezavisimyy).
2) Install the "Advanced" solution mode (Nastraivaemÿy)
129
Machine Translated by Google
field RISK
V3, V6, V7
4) In kachestve vyxodnyx znacheniy vÿbiraem znacheniya, zapisannÿe v
5) In kachestve vxodnyx znacheniy vÿbiraem znacheniya, zapisannÿe v polya
130
Machine Translated by Google
6) Select the type of check "Separation of data sets"
7) Select “Automatic definition of minimum threshold
razryadnosti ”, 3 layers of neurons, number of neurons.
131
Machine Translated by Google
Machine Translated by Google
Processing process
133
Machine Translated by Google
9) The result of the work
134
Machine Translated by Google
Machine Translated by Google
Output: In tselom pravilno reshennyx primerov (uchityvaya, chto oshibka
Do'stlaringiz bilan baham: | 
