Deep Learning

About This Course

In this course, you learn the essentials of Deep Learning. We start with a brief introduction and illustrate how to set up your software environment. We then review the foundations of artificial neural networks such as the perceptron and multilayer perceptron (MLP) networks. Next, we elaborate on convolutional neural networks illustrated with various examples. We then discuss representational learning and embeddings. Recurrent neural networks are also covered again extensively illustrated with examples. This is followed by a discussion on generative adversarial networks. The course concludes by discussing reinforcement learning.

The course provides a sound mix of both theoretical and technical insights , as well as practical implementation details. These are illustrated by several real-life case studies and examples. Throughout the course, the instructors also extensively report upon their research and industry experience.

The course features various Python Jupyter notebooks and a Python tutorial is also provided.

The course features more than 4 hours of video lectures, more than 100 multiple choice questions, and various references to background literature. A certificate signed by the instructors is provided upon successful completion.

See TBC to get a free teaser of the course contents.

We can also come and teach this course on-site in classroom format. If interested, please mail us at: Bart@BlueCourses.com.

Price

The enrollment fee for this course is EUR 250 (VAT excl.) per participant. Payments are securely handled by PayPal. If you are a company in the European Union, then we can apply VAT reverse charge. For this, please mail your VAT number to Bart@BlueCourses.com. Part of our course revenue is used towards funding organizations involvement in protecting and cleaning our oceans. See our about page to learn more about our mission statement.

Requirements

Before subscribing to this course, you should have a basic understanding of descriptive statistics (e.g., mean, median, standard deviation, histograms, scatter plots, etc.) and inference (e.g., confidence intervals, hypothesis testing). You should also have followed and completed our Machine Learning Essentials course.

Course Outline

Introduction

About
From if-then rules to deep learning
Comparison
A brief history of deep learning
What will be covered

Getting ready for deep learning

Software
Setting up your environment

Foundations of artificial neural networks

The perceptron

Concept
Activation and transfer functions
Bias

Training a perceptron

A simple iterative approach
Demo: A simple perceptron in Python
Gradient descent
The XOR problem

Multilayer perceptron (MLP) networks

Backpropagation
Automatic differentiation
Demo: Automatic differentiation in TensorFlow
Summary so far
Demo: Handwritten digits recognition with an MLP

Further aspects

Activation functions
ReLU
Initialization
Demo: The importance of initialization
Loss functions
Stochastic gradient descent
Backpropagation alternatives
Optimizers
Learning rate
Preventing overfitting
Hyperparameter optimization

Quiz

Convolutional neural networks (CNNs)

The convolutional architecture

Concept
Filters and pooling
Demo: Handwritten digits recognition with a CNN

Best practices
Dropout
Batch normalization
Data augmentation
Demo: Colored image classification with a CNN
Opening the black box
Interpretability examples with a CNN
Further aspects

Transfer learning
Demo: Transfer learning with a CNN
Variants
Demo: Locating objects with a CNN
Capsule networks
Adversarial attacks

Use cases

Artistic style transfer
Demo: Artistic style transfer example
Deep dream
Demo: Deep dreaming example

Quiz

Representational learning and embeddings

Embeddings in text

Concept
Word embeddings
word2vec
Demo: Building a word2vec model

Use cases
Generalizing embeddings
Further aspects

Variants
Demo: Graph embeddings example
Software
Categorical embeddings
Demo: Featurization with categorical embeddings

Auto-encoders

Demo: Anomaly detection with auto-encoders
Demo: Image denoising with auto-encoders

Quiz

Recurrent neural networks (RNNs)

The recurrent architecture

Concept
Demo: Text classification with an RNN
Demo: Text generation with an RNN

Further aspects

Variants
Attention and memory
Demo: Text classification with attention
Time series
Demo: Time series forecasting with an LSTM
Revisiting the CNN
Demo: Text classification with a CNN

Quiz

Generative adversarial networks (GANs)

The generative adversarial architecture

Concept
Demo: Generating digits with a GAN

Challenges
Best practices
Demo: Generating digits with a GAN, revisited
Further aspects
Variants
Use cases
Quiz

Reinforcement learning (RL)

Reinforcement learning

Concept
Q learning
Exploration versus exploitation
Demo: Q learning example
Deep Q learning
Experience replay
Demo: Deep Q learning example

Further aspects

Double deep Q learning
Demo: Double deep Q learning example
Common improvements
Variants
Software

Challenges
Quiz

Conclusions

Quiz

Course Staff

Prof.dr. Seppe vanden Broucke

Seppe vanden Broucke works as an assistant professor at the Faculty of Economics and Business, KU Leuven, Belgium. His research interests include business data mining and analytics, machine learning, process management and process mining. His work has been published in well-known international journals and presented at top conferences. He is also author of the books Beginning Java Programming (Wiley, 2015) and Principles of Database Management (Cambridge University Press, 2018). Seppe's teaching includes Advanced Analytics, Big Data and Information Management courses. He also frequently teaches for industry and business audiences. See Seppe.net for further details.