Convolutional Neural Network Walkthrough -Post 2 |Summer Research Summary

This post mentions how to understand convolutional neural network from shallow to deep.

Convolutional Neural Network

Convolutional Neural Network
Understand basic structures of convolutional neural network from stanford open-course. Pay special attention to the convolution layer’s part, and understand the convolution operation from the gif given. You may refer back to this website for more careful lookthrough later.

A Beginner’s Guide To Understanding Convolutional Neural Networks
This tutorial explains rather straightforwardly, with nice illustrations and examples. Combine with the last tutorial.
Handwritten Digit Recognition with a Back-Propagation Network
One of the groundbreaking works in pattern recognition.

Detailed Compositions

ReLU Layer - New Activation Functions
Take a look at the new ReLU activation function, compared with traditional sigmoid functions.
Pooling Layer
Pay attention to how pooling layers are backpropagated.
Dropout Layer
Another commonly-used technique to prevent overfitting in neural networks.

Visualizing Convolution
Watch the dynamic convolution illustration and several visualization of the function of individual layers.
Visualizing Convolution - YouTube Video
Visualize each convolution layers’ function and how they form a hierarchical structure.
Tensorflow Realization
This page tells how to realize CNN using tensorflow. The languages may be a little abstruse for newcomers, but at least better than nothing.

General Deep Learning Strategies

Must Know Tips/Tricks in Deep Neural Networks
The webpage shares some interesting insight into deep networks training.
How To Improve Deep Learning Performance
Some general discussion in sequential order.

Object Detection

基于深度学习的目标检测研究进展(Chinese) Introducing object detection algorithms in chronological order.