Convolutional Neural Network
الشبكة العصبية التلافيفية
Definition
A deep neural network that uses convolution layers with small filters sliding over the input to extract spatial features; used mainly for images.
Definition
A convolutional neural network is a deep network that uses convolutional layers with small learnable filters shared across spatial positions, typically followed by pooling, to learn spatial feature hierarchies from grid-like data such as images.
Practical examples
import torch
from torch import nn
cnn = nn.Sequential(
nn.Conv2d(1, 8, kernel_size=3, padding=1),
nn.ReLU(),
nn.MaxPool2d(2),
nn.Flatten(),
nn.Linear(8 * 14 * 14, 10),
)
x = torch.randn(4, 1, 28, 28) # 4 صور رمادية بحجم 28×28
print(cnn(x).shape)Expected output
torch.Size([4, 10])
Mathematical equations
S(i,j) = \sum_{m}\sum_{n} I(i+m,\, j+n)\, K(m,n)Scientific sources
References and scientific sources
-
Research paperGradient-Based Learning Applied to Document Recognition The LeNet-5 convolutional network for handwritten digit recognition.
-
Research paperDeep Residual Learning for Image Recognition Residual networks (ResNet).
-
BookDeep Learning A comprehensive reference book on deep learning, free to read on the web.
