Deep Learning

Author:Guanqun Yang

Introduction

This set of notes come from two courses
  • ECE 239AS: Neural Networks and Deep Learning
  • CS 269: Current Topics in Artificial Intelligence
Since a success of a deep learning application primarily hinges on the choice
of architecture, loss function and optimization methods, this set of notes will
be divided into sections correspond to these topics.

Loss Function

Hinge Loss

SoftMax

Optimization Methods

Gradient Descent

Architecture

Fully-Connected Neural Network (Multilayer Perceptron)