Ma’ruza rejasi

Tasvirlarda sinflashtirish masalasi

Misol uchun: 800 x 600 x 3

(3 kanalli RBG) tasvir berilgan
Ushbu holatda sinflashtirish masalasini
yechish uchun bo’lishi mumkin bo’lgan
holatlar ya’ni sinflar haqida xususiyat
qiymatlari berilgan bo’ladi (misol uchunL cat, car, airplane, dog, truck, ...)
cat

Tasvirni sinflashtirishning klassik yondashuvi

Bunda tasvirga tegishli bo’lgan yoki bo’lishi mumkin bo’lgan qirralar,

va yuqori xususiyatlar
burchaklar qiymatli aniqlanadi

Tasvirni sinflashtirish

• 
  O’quv to’plamini yig’ish
  
• 
  Mashinali o’qitish asosida sinflash bloklarini o’qitish
  
• 
  Sinf natijasini sinflash bloki yordamida bashorat qilish
  

Chuqur o’qitish algoritmlari bir-biriga bog’langan neyron tarmoq qatlamlaridan tashkil topadi

Ko’p qatlamli perceptron

Chiziqli sinflashtirish masalasida neyron tarmog’i natijasi

0.4	-0.6	1.0	1.5
1.2	1.0	-0.7	2.0
0.6	0	0.2	-0.4

212
35

5
*
+

0.8

1.1

1.0

=

-65.5

246.1

33.8

kiruvchi tasvir piksel qiymatlari
48
w(weight)
b(bias)
0-tugun qiymati
1-tugun qiymati
2-tugun qiymati
x(kirish)

0.5	0.4	1.0
0.9	1.0	-1.5

*
+

2.8

-1.2

=

-65.5

246.1

33.8

w(weight)
b(bias)
1-sinf qiymati (o’rdak)
2-sinf qiymati (mushik)
yashirin qatlam tugun qiymatlari

102.2

135.2

Machine&Deep Learning

• 
  Mashinani o’qitish bu modelni qurib olish uchun amalga oshiriladigan
  o’qitish jarayoni
  

Mashinani o’qitish

• 
  Neyron tarmog’i bu modelni tezroq va aniqroq qurib olish uchun amalga oshiriladigan jarayon
  

Neyron tarmog’i

• 
  Chuqur o’qitish bu bir nechta qatlamdan iborat bo’lgan neyron
  tarmog’ini qurish va o’qitish jarayonidir
  

Chuqur o’qitish

Chuqur o’qitish bosqichlari

Muammoni tushunib olish

Ma’lumotlarni aniqlash
Chuqur o’qitish algoritmini belgilash
Modelni o’qitish
Modelni testlash

Misol: Robot dancing

Quyidagi misolda robot raqsga tushish jarayoni chuqur o’qitish orqali amalga oshirilgan bo’lib, bunda oq’tish masalasi muhim hisoblanadi. Ushbu model 10 daqiqa davomida o’qitilganda natija yaxshi chiqmagan, 48 soat davomida o’qitish natijasida esa qurilgan model robot o’yinga tushish simulyatsiyasini ko’rsatib bergan.

Chuqur o’qitish tarmog’i turlari

• 
  Perceptron
  
• 
  Feed Forward (FF)
  
• 
  Radial Basis Network (RBN)
  
• 
  Deep Feed Forward (DFF)
  
• 
  Convolution Neural Network (CNN)
  
• 
  Recurrent Neural Network (RNN)
  
• 
  Long/Short Term Memory (LSTM)
  
• 
  Gated Recurrent Unit (GRU)
  
• 
  Auto Encoder (AE)
  
• 
  Variational AE (VAE)
  
• 
  Denoising AE (DAE)
  
• 
  Sparse AE (SAE)
  Source: https://www.guru99.com/deep-learning-tutorial.html
  

Convolutional neural networks (CNN)

CNN - bu ko'p qatlamli neyron tarmoq hamda noyob arxitekturaga ega bo’lib, har bir qatlamda ma'lumotlarning tobora murakkab xususiyatlarini chiqish uchun aniqlashga mo'ljallangan. CNN lar tanib olish masalalarini yechishda, sinflashtirish masalalarida keng foydalaniladi.

Chuqur o’qitish algoritmi asosida ishlovchi dasturlar

AI in Finance: The financial technology sector has already started using AI to save time, reduce costs, and add value. Deep learning is changing the lending industry by using more robust credit scoring. (Fintech company)
AI in HR: Under Armour, a sportswear company revolutionizes hiring and modernizes the candidate experience with the help of AI.
AI in Marketing: AI is a valuable tool for customer service management and personalization challenges. Improved speech recognition in call-center management and call routing as a result of the application of AI techniques allows a more seamless experience for customers.

Chuqur o’qitishda kuzatiladigan kamchiliklar

Data labeling - Ko'pincha zamonaviy AI modellari "nazorat ostida o'rganish“ – (supervised learning) orqali o'qitiladi. Bu shuni anglatadiki, odamlar katta hajmdagi va xatolarga olib keladishi mumkin bo'lgan asosiy ma'lumotlarni belgilashi va toifalashlari kerak.
Obtain huge training datasets - CNN kabi chuqur o’qitish usullari ba'zi hollarda tibbiyot va boshqa sohalardagi mutaxassislarning bilimlariga mos ishlaydi. Hozirgi vaqtda mashina o'qitish jarayoni nafaqat ma’lumotar aniq bo’lishini, balki yetarlicha keng va universal bo'lgan o'quv ma'lumot to'plamlarini talab qiladi.
Explain a problem - Katta va murakkab modellarni tushuntirish va unig xususiyatlarini aniqlash qiyin hisoblanadi. Bunday holatlarda aniq qarorlarni qabul qilishda xatoliklarga yo’l qo’yilishi mumkin.

Foydalanilgan adabiyotlar

Aurelian Geron, Hands on Machine Learning with Scikit-Learn Keras&Tensorflow // Second edition Concepts, Tools, and Techniques to Build Intelligent Systems, 2019, 510 pages

Primoz Potocnik, Neural Networks: MATLAB examples // Neural Networks course (practical examples)© 2012

https://www.guru99.com/deep-learning-tutorial.html

https://www.tutorialspoint.com/python_deep_learning/python_deep_learning_de ep_neural_networks.htm

https://www.mathworks.com/help/deeplearning/examples/create-simple-deep- learning-network-for-classification.html

http://hozir.org