|
Guruh magistranti Mo‘minov Suhrob 2024 y. Tekshirdi: Ochilov Mannon Musinovich
|
| bet | 4/4 | | Sana | 18.05.2024 | | Hajmi | 230,88 Kb. | | #242898 |
Bog'liq 2-amaliyKiruvchi qiymatlar
|
Og’irliklar(weights)
|
x1=2, x2=0;
b1=1, b2=3
y1=1, y2=1,5
|
I. w1=0.12, w2=0.13, w3=0.14, w4=0.15, w5=0.16, w6=0.11;
II. w7=0.13, w8=0.17, w9=0.18, w10=0.19, w11=0.2, w12=0.21, w13=0.22, w14=0.23, w15=0.24, w16=0.25, w17=0.26, w18=0.27
III. w19=0.3, w20=0.31, w21=0.32, w22=0.33, w23=0.34, w24=0.35, w25=0.36, w26=0.37, w27=0.38, w28=0,39
IV. w29=0.4, w30=0.41, w31=0.42, w32=0.43, w33=0.44, w34=0.45
|
Yuqorida keltirilgan jadvaldagi qiymatlar bo’yicha hisob kitoblarni amalga oshiramiz. Aktivatsiya funksiyasi sifatida sigmoid funksiyasidan foydalanamiz:
h1= = (2*0.12+0.15*0+1)= =0.775
h2= =
h3= =
h4= =
h5= =
h6= =
h7= =
Yuqoridagi 1-jadvalda H3(h8, h9, h10) yashirin qatlam elementlarini hisoblash uchun aktivatsiya funksiyasi sifatida ReLU funksiyasidan foydalangan holda hisoblaymiz:
h8= =
h9= =
h10= =
Yuqoridagi hisob-kitoblar bajarilgandan so’ng O (y1, y2) to’plam uchun aktivatsiya funksiyasi sifatida Softmax funksiyasi funksiyasidan foydalangan holda hisoblaymiz:
y1= =
y2= =
Hisob-kitoblarni bajarib bo’lgach ushbu neyron tarmoq uchun Multi-class cross-entropy/categorical cross-entropy asosida yo‘qotilish (Loss) qiymatini hisoblanadi. Ushbu qiymat quyidagi formula asosida hisoblanadi:
Kiruvchi qiymatlar
|
Chiquvchi qiymatlar
|
x1=2, x2=0
y1=1, y2=1,5
|
y1=0.49, y2=0.50
|
|
| |
