Week 13- Optimization Algorithms

Click here to read about Week 13.

Welcome to Week 13. During this week, we will turn our attention to non-convex problems and methods for solving them. In general, we can only guarantee convergence to a local minimum, but there are several algorithms that can find a very advantageous local minimum in the context of machine learning problems like training neural networks. The focus will be on gradient descent, and improvement which handles large condition numbers better called gradient descent with momentum, and adaptive algorithms which are often the default choice for training neural networks such as ADAM.

Your final homework assignment is due May 18th at the latest.

Here are more details on what you need to do this week:

• Go through module 11 and complete the week 13 readings.
• Attend our second meetup on Monday at 6:45 p.m. ET on Zoom
• Begin working on Lab 7 (available soon)