|
Kompyuter tizimlari
|
bet | 3/5 | Sana | 01.01.2023 | Hajmi | 0.85 Mb. | | #37392 |
Bog'liq 3012107720, notification-file, application-file, 1669973412 (3), 1669120852, 1671794695, 1671786083, 1671627717, 6-Hhg2maExef6D4dssx4y3oBHURCKfsq, AgioGbFzDYdNWpPFYeiuNAhafTAYCWxy, 1, Axborot texnologiyalari va kommunikatsiyalarini rivojlantirish v-www.hozir.org, - Raspberry Pi for Beginners Revised Edition 2014 (2011), electronics-10-00115-v3Tajriba SVM va
MLP neyron tarmog'i uchun biz performRecognition.py yordamida
tasniflagichni sinab ko'ramiz. Ushbu skriptda biz 14-rasmda ko'rsatilganidek, avval saqlagan faylni tasniflaymiz.
NATIJA
KNN uchun 1 dan 15 gacha bo'lgan k qiymati uchun natija bir xil natijaga ega bo'ladi, bu 99,26% ni tashkil etadi, bu 18-rasmda ko'rsatilganidek, eng yuqori aniqlikdir.
14-rasm - Chaqiruv klassifikatori fayli
SVM tasnifi natijasi uchun quyidagi 21-rasmda ko'rsatilgan.
rasm - Klassifikatorning aniqligi
Sinov ma'lumotlarini baholash quyidagi 19-rasmda ko'rsatilgandek natija beradi.
rasm - Sinov ma'lumotlarini baholash
rasm - SVM klassifikatori natijasi
MLP neyron tarmog'ining tasnifi natijasi uchun
quyidagi 22-rasmda ko'rsatilgan.
rasm - MLP neyron tarmog'i natijasi
rasmda ko'rsatilganidek, o'quv ma'lumotlar to'plamidan besh (5) tasodifiy tasvirga ega tekshiruv klassifikatori uchun.
XULOSA
Natijadan ko'rinib turibdiki, KNN va SVM ma'lumotlar to'plamini to'g'ri bashorat qiladi, lekin MLP Neyron Network uchun bu 9 raqamini taxmin qilishda xatodir. Buning sababi, KNN va SVM uchun u to'g'ridan-to'g'ri xususiyatni ajratib olish orqali bashorat qiladi. Ammo MLP uchun bu chiziqli bo'lmagan funktsiyadir. Shunday qilib, chiziqli bo'lmagan modellarni o'rganish
uchun ko'proq mos keladi. Yashirin qatlamlarga ega MLP esa bir nechta mahalliy minimal mavjud bo'lsa, qavariq bo'lmagan yo'qotish funktsiyasiga
ega. Shuning uchun turli xil tasodifiy og'irlikni ishga tushirish har xil tekshirish aniqligiga olib kelishi mumkin. Ammo u Keras bilan konvolyutsion neyron tarmoqlardan foydalanish orqali yaxshilanishi mumkin.
ADABIYOTLAR
[1] Youssouf Chherawala, Partha Pratim Roy and Mohamed Cheriet, “Feature Set Evaluation for Offline Handwriting Recognition Systems: Application to the Recurrent Neural Network,” IEEE TRANSACTIONS ON CYBERNETICS, VOL. 46, NO. 12, DECEMBER 2016
[2] Nurul Ilmi, Tjokorda Agung Budi W and Kurniawan Nur R, “Handwriting Digit Recognation using Local Binary Pattern Varience and K-Nearest Neighbor,” 2016 Fourth International Conference on Information and Communication Technologies (ICoICT).
[3] Tobias Kutzner, Mario Dietze, Ingrid Bönninger, Carlos M. Travieso, Malay Kishore Dutta, and Anushikha Singh, “Online Handwriting Verification with Safe Password and Incresing Number of Features,” 2016 3rd International Conference on Signal Processing and Integrated Networks (SPIN).
|
| |