Genetik algoritm va uning qo’llaniladigan masalalar
![Buning uchun dastlabki ma’lumotlarni tayyorlashni bajarish zarur. Bunday
dastlabki tayyorgarlik juda oddiy bо‘lishi mumkin, - masalan, barcha belgilarning
qiymatlarini masshtablash protsedurasi yordamida (0,1) intervalga о‘tkazish, balki
ancha murakkab algoritmlardan foydalanish ham mumkin. Biroq bu yerda shuni
ta’kidlash kerakki, neyron tо‘rini qurishdagi uzoq muddatli maqsad masalani
yechimini olish jarayoniga tо‘g‘ridan-tо‘g‘ri ta’sir etishni minimallashtirishdan
iborat, chunkineyron tо‘ralrining asosiy ustunligi о‘zining yechimlarini ishlab
chiqishning potensial mavjudligi, imkoniyati. Amaliyotda juda yaxshi yechimlar
olinadi, qachonki, mutaxassis, kо‘rilayotgan muammoni bilimlar sohasida aniq
tasavvur qilaolsa va qurilgan neyron tо‘rining konseptual muammolarini tushunib
yetgan bо‘lsa.
Neyron tо‘rini о‘qitish uchun foydalaniladigan ma’lumotlar odatda ikki
kategoriyaga bо‘linadi: bir inchi kategoriya ma’lumotlari о‘qitish uchun ishlatiladi,
ikkinchi kategoriya ma’lumotlari esa tо‘rni testlash uchun ishlatiladi. Neyron
tо‘rining haqiqiy sifati faqat tо‘rni testlash vaqtida namoyon bо‘ladi, tо‘rni
о‘qitishning muvaffaqiyatli tugaganligi testlash jarayonida notо‘g‘ri ishlash
belgilarining yо‘qligidadir. Testlash jarayoni shunday tashkil qilish kerakki, tо‘rni
ishlash jarayonida olingan bilimlarni umumlashtirish qobiliyati baholananishi
zarur. Bu yerda bilimlarni umumlashtirish deganda berilgan ma’lumotlar bо‘yicha
masalani tо‘g‘ri yechimini ishlab chiqish qobiliyatiga aytiladi.
Neyron tо‘rini о‘qitishni oddiy misolda kо‘rib chiqamiz. Aytaylik neyron
tо‘ri bitta kirish va bitta chiqish elementidan tarkib topgan. Maktab kursidan
ma’lumki fizika darslarida tajribalar о‘tkaziladi va dekart tekisligida uning grafigi
quriladi. Masalan, quyidagi tajriba ma’lumotlari olingan: kirish signallarini qiymati
X={xi}, Y={yi}, i=1,2,….,n. Grafikda korrelyatsiya maydonidagi nuqtalarning
joylashganligiga qarab bu yerdagi kirish va chiqish orasidagi analitik bog‘lanish
tо‘g‘ri chiziqga yaqinligshi kо‘rinib turibdi [44-48].
11](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_10.png?v=1)
![o'lchamlari turli darajadagi abstraktsiyani ta'minlashi mumkin.) "Chuqur o'rganish"
tarkibidagi "chuqur" so'zi ma'lumotlarning o'zgarishi qatlamlari sonini anglatadi.
Aniqrog'i, chuqur o'rganish tizimlari muhim ahamiyatga ega kredit berish yo'li
(CAP) chuqurligi. CAP - bu kirishdan chiqishga transformatsiyalar zanjiri. CAP-
lar kirish va chiqish o'rtasidagi potentsial sababiy aloqalarni tavsiflaydi. A
feedforward neyron tarmoq, CAPs chuqurligi tarmoqqa tegishli va maxfiy
qatlamlar soni va plyus bitta (chunki chiqish qatlami ham parametrlangan). Uchun
takrorlanadigan neyron tarmoqlari, unda signal qatlam orqali bir necha marta
tarqalishi mumkin, CAP chuqurligi potentsial cheksizdir.[2] Hech qanday umumiy
kelishilgan chuqurlik chegarasi sayoz o'rganishni chuqur o'rganishdan ajratmaydi,
ammo aksariyat tadqiqotchilar chuqur o'rganish uchun CAP chuqurligi 2 dan
yuqori ekanligini o'z ichiga oladi degan fikrga qo'shilgan. 2-chuqurlikdagi CAP har
qanday funktsiyani taqlid qilishi mumkinligi nuqtai nazaridan universal taxminiy
vosita sifatida ko'rsatilgan. Bundan tashqari, ko'proq qatlamlar tarmoqning
funktsional taxminiy qobiliyatiga qo'shilmaydi. Chuqur modellar (CAP> 2) sayoz
modellarga qaraganda yaxshiroq xususiyatlarni chiqarishga qodir va shuning
uchun qo'shimcha qatlamlar funktsiyalarni samarali o'rganishda yordam beradi.
Chuqur o'rganish me'morchiligini a bilan qurish mumkin ochko'z qatlam-qavat
usuli. Chuqur o'rganish ushbu abstraktsiyalarni ajratishga va qaysi xususiyatlar
ishlashni yaxshilaydiganligini tanlashga yordam beradi. Uchun nazorat ostida
o'rganish vazifalar, chuqur o'rganish usullari yo'q qilinadi xususiyati muhandislik,
ma'lumotlarga o'xshash ixcham oraliq vakolatxonalarga tarjima qilish orqali asosiy
komponentlarva vakolatxonada ortiqchalikni olib tashlaydigan qatlamli
tuzilmalarni olish. Chuqur o'rganish algoritmlarini nazoratsiz o'qitish vazifalariga
qo'llash mumkin. Bu muhim foyda, chunki yorliqsiz ma'lumotlar yorliqli
ma'lumotlarga qaraganda ancha ko'p. Nazorat qilinmaydigan tarzda o'qitilishi
mumkin bo'lgan chuqur tuzilmalarga misollar asab tarixi kompressorlari va chuqur
e'tiqod tarmoqlari.
17](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_16.png?v=1)
![FOYDALANILGAN ADABIYOTLAR
1. Cybenko, G. 1989. Approximation by superpositions of a sigmoidal
function Mathematics of Control, Signals, and Systems, 2(4), 303–314.
2. Linnainmaa, Seppo (1970). The representation of the cumulative
rounding error of an algorithm as a Taylor expansion of the local rounding errors
(Masters) (in Finnish). University of Helsinki. pp. 6–7.
3. Kelley, Henry J. (1960). "Gradient theory of optimal flight paths".
ARS Journal. 30 (10): 947–954. doi:10.2514/8.5282.
4. Rosenblatt, Frank. x. Principles of Neurodynamics: Perceptrons and
the Theory of Brain Mechanisms. Spartan Books, Washington DC, 1961
5. Werbos, Paul (1982). "Applications of advances in nonlinear
sensitivity analysis" (PDF). System modeling and optimization. Springer. pp. 762–
770. Archived (PDF) from the original on 14 April 2016. Retrieved 2 July 2017.
6. Rumelhart, David E., Geoffrey E. Hinton, and R. J. Williams.
"Learning Internal Representations by Error Propagation". David E. Rumelhart,
James L. McClelland, and the PDP research group. (editors), Parallel distributed
processing: Explorations in the microstructure of cognition, Volume 1:
Foundation. MIT Press, 1986.
7. Hastie, Trevor. Tibshirani, Robert. Friedman, Jerome. The Elements
of Statistical Learning: Data Mining, Inference, and Prediction. Springer, New
York, NY, 2009.
8. Rosenblatt, Frank (1958). "The Perceptron: A Probabilistic Model For
Information Storage And Organization in the Brain". Psychological Review. 65
(6): 386–408. CiteSeerX 10.1.1.588.3775. doi:10.1037/h0042519. PMID
13602029. S2CID 12781225.
9. Rosenblatt, Frank (1962). Principles of Neurodynamics. Spartan, New
York.
10. Schmidhuber, Juergen (2022). "Annotated History of Modern AI and
Deep Learning". arXiv:2212.11279 [cs.NE].
26](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_25.png?v=1)
![ILOVA
Genetik algoritmlarni o'rganish, Python dasturlash tili bilan amalga
oshirilishi mumkin. Ushbu misol kod, genetik algoritmni boshlang'ich
populyatsiyani optimallashtirish masalasini hal qilish uchun yaratilgan oddiy bir
skriptni ko'rsatadi:
import numpy as np
def objective_function(x):
"""Maqsad funksiyasi."""
return sum(x*2)
def create_population(population_size, chromosome_length):
"""Boshlang'ich populyatsiyani yaratish."""
return np.random.rand(population_size, chromosome_length)
def calculate_fitness(population):
"""Populyatsiya uchun fitness qiymatlarni hisoblash."""
return np.apply_along_axis(objective_function, 1, population)
def selection(population, fitness):
"""Tanlov (Selection) jarayoni."""
selected_indices = np.argsort(fitness)[:len(fitness)//2]
return population[selected_indices]
def crossover(parents):
"""Qo'llanish (Crossover) jarayoni."""
crossover_point = np.random.randint(1, len(parents[0]))
children = []
for i in range(0, len(parents)-1, 2):
27](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_26.png?v=1)
![child1 = np.concatenate((parents[i][:crossover_point], parents[i+1]
[crossover_point:]))
child2 = np.concatenate((parents[i+1][:crossover_point], parents[i]
[crossover_point:]))
children.extend([child1, child2])
return np.array(children)
def mutation(population, mutation_rate):
"""Mutatsiya (Mutation) jarayoni."""
mutated_indices = np.random.rand(*population.shape) < mutation_rate
population[mutated_indices] = np.random.rand(sum(mutated_indices))
return population
def genetic_algorithm(population_size, chromosome_length, generations,
mutation_rate):
"""Genetik algoritm."""
population = create_population(population_size, chromosome_length)
for _ in range(generations):
fitness = calculate_fitness(population)
selected_population = selection(population, fitness)
children = crossover(selected_population)
mutated_population = mutation(children, mutation_rate)
# Populyatsiyani yangilash
population[:len(mutated_population)] = mutated_population
# Eng yaxshi natijani chiqarish
best_individual = population[np.argmin(fitness)]
best_fitness = objective_function(best_individual)
28](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_27.png?v=1)
![print(f"Iteration {_+1}/{generations}, Best Fitness: {best_fitness}")
return best_individual
# Genetik algoritmni boshqarish
population_size = 100
chromosome_length = 5
generations = 50
mutation_rate = 0.1
best_solution = genetic_algorithm(population_size, chromosome_length,
generations, mutation_rate)
print("Eng yaxshi hal:", best_solution)
children = crossover(selected_population)
mutated_population = mutation(children, mutation_rate)
# Populyatsiyani yangilash
population[:len(mutated_population)] = mutated_population
# Eng yaxshi natijani chiqarish
best_individual = population[np.argmin(fitness)]
best_fitness = objective_function(best_individual)
print(f"Iteration {_+1}/{generations}, Best Fitness: {best_fitness}")
return best_individual
# Genetik algoritmni boshqarish
population_size = 100
chromosome_length = 5
29](https://docx.uz/documents/3559641c-2356-4842-a97d-840428466184/page_28.png?v=1)
GENETIK ALGORITM VA UNING QO’LLANILADIGAN MASALALAR.docx