Mastering Deep Learning [23-24 November Brisbane/Australia]

For up-to-date Information, testimonials, updated session's information, visit: 

http://www.datasmelly.com/trainingDL_2.html

Outline

This comprehensive two day course provides you with a complete introduction to Deep Learning.

Deep Learning has revolutionized analytics in just over five years. The field itself is changing very quickly, with interesting developments every day. This workshop is aimed to give you a complete introduction to Deep Learning. At the end of the workshop, you will have (a) a better understanding/appreciation of deep learning, (b) a better understanding of problem-domains that can be solved by DL, (c) build deep learning solution using TensorFlow. 
Main topics include, Artificial Neural Networks (ANN), Deep ANN, AutoEncoders, Deep Belief Networks, Convolutional Neural Networks, Recurrent Neural Networks, LSTM, Attention-based methods, Generative Adversarial Networks (GANs), Variational Auto-encoders, Deep Reinforcement Learning, etc.

Why you should attend? You will learn:

How Facebook tags you and your friends in pictures?

How Google translates from English to Chinese?

How AlphaGo learned to play Atari games beating humans?

How Chatbots learn to mimic a human?

How can you do superior time series analysis?

Etc. Etc.

Motivations

Outside academia, not many comprehensive Deep Learning courses exists. This course is aimed to bridge the gap between current Deep Learning (research and developments) and other stake-holders that can greatly benefit from Deep Learning techniques.

What this course is not?

This workshop is not a training of the use of Deep Learning libraries. We expect some very basic understanding of Machine Learning, however, some proficiency with Python or R programming languages. This workshop is training in underlying deep learning methodologies, fundamental concepts and algorithms that every data scientists should be familiar with. Of course, we illustrate and demonstrate all concepts in TensorFlow (with Python Jupyter notebooks).

Duration

2 days (9 + 9 hours)

Training Breakdown

(*) Note the unit breakdown in the following is likely to change, but it provides a general overview of the topics that this training will cover.

Module

Session 1 -- Introduction

• Introduction to Machine Learning
  • Linear/Logistic Regression
  • Optimization
    • Gradient Descent, Stochastic Gradient Descent
  • Bias-Variance Trade-off, Under-fitting/Over-fitting
  
  
  • Need for low-bias vs. low-variance Models in ML
  
  
• Artificial Neural Networks
  
  
  • Gradients
  
  
  • Regularization -- Dropout
  
  
  • Model Architecture
  
  
• Deep Learning


• Representation Learning

Session 2 -- Auto-Encoders and Deep Belief Networks

• Unsupervised ANN


• Restricted Boltzmann Machines (RBM)


• Auto-Encoders (AE)
  
  
  • Principal Component Analysis (PCA)
  
  
  • Singular-value-Decomposition (SVD)
  
  
  • AutoEncoders Architecture
  
  
• Sparse Auto-encoders


• Deep Belief Nets

Session 3 -- Convolutional Neural Networks (CNN)

• Introduction
  
  
  • Convolution
  
  
  • Feature Maps, Max Pooling
  
  
  • Batch Normalization
  
  
• CNN Architectures


• Face Recognition
  
  
  • Triplet Loss
  
  
• CNN for non-images

Session 4 -- Recurrent Neural Networks (RNN)

• Introduction
  
  
  • Various Architectures
  
  
  • Applications
  
  
• RNN Embeddings


• LSTM, GRU


• Attention-based models
  
  
  • Transformers
  
  
• Building Chatbots

Session 5 -- Generative Neural Networks

• Introduction


• Generative Adversarial Networks


• Variational Auto-Encoders

Session 6 -- Deep Reinforcement Learning

• Introduction
  
  
  • Learning Optimal Policies
  
  
• Q-Learning
  
  
  • Deep Q-Learning (DQN)
  
  
  • Double Deep Q-Learning (DDQN)
  
  
• Policy-based Methods
  
  
  • Policy-gradient Method
  
  

Session 7 -- Networking

Salient Features

Comprehensive and State-of-the-art training in Deep Learning

Small Group -- Max 15. Opportunity to meet and mingle

Frequent Questions

Why should I attend this course?

Good question! Let us ask you some counter questions.

• • 1) Are you interested in exploring Deep Learning with some breath and depth?

• • 2) Are you curious about the inner workings of deep learning algorithms?

• • 3) Do you want to understand how Google Translate works? Converting English into Flemish, Chinese into Afrikaans, etc. 

• • 4) Working out on a chatbot technology -- trying to understand how it works? 

• • 5) Curious about inner working of reinforcement learning, and how alpha-Go learns to play Atari games better than humans? 

• • 5) How computers recognize faces or Facebook tags you automatically in an image? 

If answer to any of the above questions is yes, then you should attend this course.

How is this course different from others?

There are not many courses on Deep Learning, most are offered as part of post-graduate degree or diploma by universities. Non-academic units have a too narrow focus on certain technologies, for example, 'Deep Learning with Theano or Torch'. The extra layer of technology around the algorithms confuses the underlying message.
We have designed this course around the core concepts and fundamental principles -- conveyed in a total technology agnostic way. The underlying concepts are taught through mathematical notations.

Do I need a deep Math background?

Not deep. But you should be familiar with some basic statistics, Linear Algebra and Calculus. We will try our best to explain the foundations, but sometimes, time may not permit. We will post some foundation slides on 'Reading List' of this web page.

Do I need to bring my laptop?

Yes, there is a lab component in most of the lectures. It will be beneficial if you bring your laptop, to run this code on your computer.

Do I need to know computer programming?

Yes. Some computer programming experience is required if you are interested in implementing some of the ideas in the workshop.

I am not a Python programmer?

That is absolutely fine. Keep checking the 'Reading List' section on this web page -- we will share some 'Introduction to Python' lectures. If you are an R/Matlab programmer, we encourage you to get yourself familiarize with Python syntax before coming to the workshop.

What does course-kit consist of?

• • 1) Book consisting of printed slides (over 300 pages)

• • 2) Certificate of Attendance (posted in 2-3 weeks after attendance), if requested

• • 3) Welcome pack

Who will deliver this unit?

This course will delivered by a 'Principle Data Scientist' (highest designation at DataSmelly) -- the coordinator will have a Ph.D in Machine Learning or related area and over six years of research and development experience.