Convolutional Neural Networks

October 11, 2026 deep-learning

Handwritten notes on CNNs: convolution operations, pooling layers, feature maps, architectures (LeNet, AlexNet, VGG, ResNet), and transfer learning.

These notes cover the architecture and intuition behind convolutional neural networks:

Convolution Operations: Filters, kernels, stride, padding, and how feature maps are computed.
Pooling Layers: Max pooling, average pooling, and spatial dimension reduction.
Classic Architectures: LeNet, AlexNet, VGG, GoogLeNet, ResNet — evolution of CNN design and skip connections.
Transfer Learning: Pre-trained models, fine-tuning strategies, and feature extraction for downstream tasks.