Deep Learning Web Course
Welcome to Math 452
contents
Module 0 Get started: course information and preparations:
Course information, requirements and reference
Course background and introduction
Introduction to Python and Pytorch
Preliminary Quiz
Module 1: Linear machine learning models
Machine learning basics, popular data sets
Linearly separable sets
Logistic regression
KL-divergence and cross-entropy
Support vector machine and relation with LR
Optimization and gradient descent method
Homework 1
Module 1 Programming Assignment
Quiz 1
Module 2: Probability and training algorithms
Introduction to probability
Probabilistic derivation of logistic regression models
Convex functions and convergence of gradient descen
Stochastic gradient descent method and convergence theory
MNIST: training and generalization accuracy
Homework 2
Week 2 Programming Assignment
Quiz 2
Module 3: Deep neural networks
Nonlinear models
Polynomials and Weierstrass theorem
Finite element method
Deep neural network functions
Universal approximation properties
Application to data classification
DNN for image classification
Monte Carlo Methods
Building and Training Deep Neural Networks (DNNs) with Pytorch
Homework 3
Week 3 Programming Assignment
Quiz 3
Module 4: Convolutional neural networks
Convolutional neural networks
Convolutional operations on images
Some classic CNN
Training CNN with GPU on Colab
Building and Training Convolutional Neural Networks (CNNs) with Pytorch
Homework 4
Week 4 Programming Assignment
Quiz 4
Module 5: Normalization, ResNet and Multigrid
Data normalization and weights initialization
Batch normalization
Building and Training ResNet with Pytorch
Multigrid Method for Finite Element
Homework 5
Week 5 Programming Assignment
Quiz 5
Module 6: MgNet
MgNet: a special CNN based on multigrid method
Multigrid and MgNet
Multigrid and MgNet
MATH 452: Final Project
.ipynb
.pdf
Binder
Live Code
Contents
Download the Module 1 homework here: Homework1
Homework 1
¶
Download the Module 1 homework here:
Homework1
¶
previous
Optimization and gradient descent method
next
Module 1 Programming Assignment
