M08 · Optimization
A free, illustrated L2 to L3 course of 11 modules in the Mathematics track. Modules include: Optimization problems and their geometry; Convex sets and convex functions; Gradient descent and step sizes.
M08 L2 to L3 · Intermediate to Advanced 11 modules · about 44 lessons Planned · Ring 3
Part of the Mathematics track. Ships in Ring 3 (Breadth) of the roadmap.
Modules
- Optimization problems and their geometry
- Convex sets and convex functions
- Gradient descent and step sizes
- Momentum, Nesterov and adaptive methods
- Newton and quasi-Newton methods
- Stochastic gradient descent and variance
- Constrained optimization and Lagrange multipliers
- KKT conditions and duality
- Linear and quadratic programming
- Proximal methods and non-smooth problems
- Non-convex landscapes: saddles, plateaus and minima
Signature illustrations
The pictures a learner should still see in their head a year later:
Optimizer paths on contour plots, convexity chord test, Lagrange tangency picture, duality gap diagram, escaping a saddle.
Status
This course is planned, not written yet. It ships with Ring 3 (Breadth).
Every lesson will follow the same twelve-part blueprint with at least three purpose-made figures. Watch the repository to hear when modules ship, or help write and review it.