What is this video about, and why is it important?

This video will serve as an introduction to PyTorch, a dynamic, deep learning framework in Python. In this video, you will learn to create simple neural networks, which are the backbone of artificial intelligence. We will start with fundamental concepts of deep learning (including feed forward networks, back-propagation, loss functions, etc.) and then dive into using PyTorch tensors to easily create our networks. Finally, we will CUDA render our code in order to be GPU-compatible for even faster model training.

What you’ll learn—and how you can apply it

• Deep learning basics and you can apply it to your domain (X + AI)
• PyTorch platform basics and you can apply it to any deep learning problem
• CUDA rendering, which will allow you to train your networks very quickly

This video is for you because…

• You may be an experienced AI researcher (academia or industry) with years of experience, and may have coded in platforms such as TensorFlow and Theano before, but may be a bit hesitant to transition into PyTorch. This introductory video will show you how easy it is to switch and the benefits you will reap with PyTorch’s dynamic nature.
• You may also be a software engineer or computer science student or enthusiast looking to get started with deep learning. For you, PyTorch is the best platform to start with because of its simple, yet powerful interface. It makes implementing deep networks very transparent, which allows you to validate all the mathematical concepts you are learning. Familiarity with basic deep learning concepts is preferred but not required as we will cover the math behind the code as well.

Prerequisites:

• An understanding of algebra and basic calculus
• Basic python skills (knowledge of functions, classes, etc.)

Materials or downloads needed in advance:

• Download and install PyTorch (Instructions provided in the forthcoming GitHub repo)
• Download corresponding Jupyter notebooks via forthcoming GitHub repo