Deep Learning (with TensorFlow)

Facilitated by the intersection of inexpensive computing power, unprecedentedly large data sets, and clever computational statistics advances, Deep Learning algorithms are driving an Artificial Intelligence revolution. Deep Learning has emerged as uniquely influential across a broad range of the statistical domain, including classification (e.g., visual recognition, sentiment analysis), prediction (e.g., stock markets, health outcomes) and generation (e.g., creating images, composing music). In the past few years, Deep Neural Networks have made their way into countless everyday applications, including Tesla’s Autopilot, Siri’s voice recognition, Facebook’s face identification, and hundreds of products at Google such as Inbox with its suggested replies. Indeed, Deep Learning algorithms have exceeded human performance on previously intractable computational problems like language translation, object detection, and the game of Go.

Via analogy to biological neurons and human perception, this course is an introduction to artificial neural networks that brings high-level theory to life with interactive labs featuring TensorFlow, the most popular open-source Deep Learning library. Essential theory will be covered in a manner that provides students with an intuitive understanding of Deep Learning’s underlying foundations. Paired with hands-on code run-throughs in Jupyter notebooks as well as strategies for overcoming common pitfalls, this foundational knowledge will empower individuals with no previous understanding of neural networks to build production-ready Deep Learning applications across the major contemporary families: Convolutional Nets for machine vision; Long Short-Term Memory Recurrent Nets for natural language processing and time series analysis; Generative Adversarial Networks for producing realistic images; and Reinforcement Learning for playing video games.


Goals
This is a “short course” of five weeks, with six hours of class per week. Classes will be given in a lab setting, with hands-on exercises mixed with lectures. Students should bring a laptop to class. You will have the option of creating and completing your own major Deep Learning project over the duration of the program.

Who Is This Course For
This course is perfect for software engineers, data scientists, analysts, and statisticians with an interest in Deep Learning. No previous knowledge of Deep Learning is assumed. Previous experience with statistics or machine learning is not necessary, but will be an asset if you have it. Code examples are provided in Python, so familiarity with it or another object-oriented programming language would be helpful.

Prerequisites
It will be challenging to follow along through the code demos and exercises without some experience in:

• Object-oriented programming, ideally Python 
• Simple shell commands, e.g., in Bash (tutorial of the fundamentals)

Outcomes
By the end of the course, you will be able to:

• Build Deep Learning models in TensorFlow and Keras
• Understand the language and theory of Artificial Neural Networks
• Excel across a broad range of computational problems including Machine Vision, Natural Language Processing and Deep Reinforcement Learning
• Create algorithms with state-of-the-art performance by fine-tuning model architectures
• Self-direct and complete your own Deep Learning project