About this course

In this course, you will begin by understanding the foundations of PyTorch for deep learning. You’ll explore how to calculate and implement logistic regression using PyTorch for binary classification tasks.

Next, you’ll dive into SoftMax regression and how it’s used for multi-class classification. By learning to implement the SoftMax function in PyTorch, you will gain an understanding of how to apply SoftMax regression to multi-class problems and fine-tune models accordingly.

The course then shifts to shallow neural networks, where you will explore how neural networks function with a single input layer and gradually progress to networks with multiple hidden layers. Through practical exercises, you will implement backpropagation and activation functions, which are crucial for effective learning in neural networks.

Deep neural networks are introduced, and you’ll learn to design more complex architectures. Techniques such as dropout for regularization, the role of weight initialization, and momentum in gradient descent will be covered. Additionally, batch normalization will be explored to stabilize and accelerate training.

Next, you will study convolutional neural networks (CNNs), which are especially powerful for image classification. You will implement CNNs using PyTorch, learning to work with convolution layers, pooling layers, and multiple input-output channels. Furthermore, you’ll explore how to integrate GPUs for faster computation and take advantage of PyTorch’s torch vision models.

The course culminates in a hands-on final project where you’ll apply the concepts you’ve learned by building a classification model for the Fashion MNIST dataset. This project will allow you to put all the skills into practice, from model development to fine-tuning and evaluation.

What you will learn:

After completing this course, you will be able to:

Explain the concept of logistic regression, cross-entropy loss, and the SoftMax function.
Apply PyTorch for implementing logistic regression, SoftMax regression, and neural network models.
Analyze the role of activation functions, backpropagation, and regularization techniques like dropout in deep neural networks.
Evaluate the effectiveness of deep learning architectures, including shallow and convolutional neural networks, in various machine learning tasks.
Create and optimize complex neural networks for classification tasks, utilizing techniques such as batch normalization and gradient descent with momentum.

Course Syllabus

Module 1: Logistic Regression Cross Entropy Loss

Welcome

Video: Course Introduction
Reading: Course Introduction
Reading: Course Overview
Helpful Tips for Course Completion
Learning Objectives and Syllabus
Grading Scheme

Logistic Regression Cross Entropy Loss

Logistic Regression Cross Entropy Loss
Logistic Regression Mean Square Error
Logistic Regression Cross Entropy
Module-level Graded Quiz: Logistic Regression Cross Entropy Loss

Module 2: Softmax Regression

Softmax Prediction

Softmax

Softmax function

Softmax function: Using Lines to classify data Prediction

Softmax PyTorch

Softmax PyTorch
Softmax Classifier 1
Softmax Classifier 2
Module-level Graded Quiz: Softmax Regression

Module 3: Shallow Neural Networks

Neural Networks in One Dimension

What’s a neural network?
Neural Networks in One Dimension
Graded Quiz:Neural Networks

Neural Networks More Hidden Neurons

More Hidden Neurons
More Hidden Neurons

Neural Networks with Multiple Dimensional Input

Neural Networks with Multiple Dimensional Input
Multi-Dimensional Neural Networks

Multi-Class Neural Networks

Multi-Class Neural Networks
Multi-Class Neural Networks with MNIST

Backpropagation

Backpropagation

Activation Functions

Activation Functions
Activation Functions
Neural network with different Activation Functions
Module-level Graded Quiz:Multi-Class Neural Networks

Module 4: Deep Networks

Deep Neural Networks

Deep Neural Networks
Deep Neural Networks
Graded Quiz: Deep Neural Networks
Deeper Neural Networks: nn.ModuleList()
Deeper Neural Networks : nn.ModuleList()

Dropout

Dropout
Dropout Classification
Dropout Regression

Neural Network Initialization Weights

The Role of Dropout in Regularization and Model Generalization
Neural Network Initialization Weights
Initialization with Same Weights
Initialization Xavier
Understanding the Importance of Initialization Techniques in Deep Learning

Gradient Descent with Momentum

Gradient Descent with Momentum
Momentum with Different Polynomials
Neural networks with momentum

Batch Normalization

Batch Normalization
Batch Normalization with the MNIST Dataset
Graded Quiz: Batch Normalization
Module-level Graded Quiz:Deep Networks

Module 5: Convolutional Neural Networks

Convolution

Convolution
What's Convolution

Activation Functions and Max Pooling

Activation Functions and Max Pooling
Activation Functions and Max Pooling

Multiple Input and Output Channels

Multiple Input and Output Channels
Multiple Input and Output Channels

Convolutional Neural Network

Convolutional Neural Network
Convolutional Neural Network for MNIST
Convolutional Neural Network Simple example
Convolutional Neural Network MNIST
Convolutional Neural Networks with Batch Norm (Honors only)

Torch Vision Models

Torch Vision Models

Graphics Processing Unit

Graphics Processing Unit
GPU Setup
Module-level Graded Quiz: Convolutional Neural Networks

Module 6: Final Project

Final Project

Practice Project
Final Project
Fashion MNIST Classification Assignment

Course Wrap-up

Congratulations and Next Steps
Thanks from the Course Team

General Information

This course is self-paced.
This platform works best with current versions of Chrome, Edge, Firefox, Internet Explorer, or Safari.

Recommended Skills Prior to Taking this Course

The following skills are required to be successful with this course:

High School level Mathematics
Working knowledge of Python.
Machine Learning with Python
Fundamentals of Deep Learning
Working knowledge of PyTorch

PyTorch for Deep Learning

Language

Topic

Skills You Will Learn

Offered By

Estimated Effort

Platform

Last Update