I have some great news. Thanks In no small part to the strong support of all of you who have enrolled in courses, I’ve been able to accelerate the course production schedule. Course 313, Advanced Neural Network Methods, is complete two months ahead of schedule.
The course has six major sections covering the topics you need to bring basic neural networks up to high levels of performance: Regularization, Dropout, Skip-layers, Computation Graphs, Optimizers, and Initialization. It also shows how these apply in an autoencoder while processing images of the surface of Mars.
The course is based on the Cottonwood machine learning framework and demonstrates how to freely experiment with deep learning concepts of your own.
Thank you and for your enthusiasm and dedication working through the courses! Your experiences give me great ideas for how to make the material even more helpful and motivation to keep it coming at a good pace. I couldn’t do it without you.
I love solving puzzles and building things. Machine learning lets me do both. I got started by studying robotics and human rehabilitation at MIT (MS '99, PhD '02), moved on to machine vision and machine learning at Sandia National Laboratories, then to predictive modeling of agriculture DuPont Pioneer, and cloud data science at Microsoft. At Facebook I worked to get internet and electrical power to those in the world who don't have it, using deep learning and satellite imagery and to do a better job identifying topics reliably in unstructured text. Now at iRobot I work to help robots get better and better at doing their jobs. In my spare time I like to rock climb, write robot learning algorithms, and go on walks with my wife and our dog, Reign of Terror.
Preview1.0 How Regularization Works
Start1.1 Constructing a Network with Regularization (7:14)
Start1.2 Executing Regularization during Stochastic Gradient Descent (3:35)
Start1.3 L1 Regularization (LASSO) (4:45)
Preview1.4 L2 Regularization (Ridge or Tikhonov) (2:15)
Start1.5 Regularization in Action (2:55)
Start1..6 Custom Regularizers (3:25)