Optimization and learning
See gradients move a state, investigate bad step sizes, and connect controlled updates to the project’s optimization records.
What you will be able to do
Trace an update, test its objective, and explain the limits of convergence and reproducibility claims.
Open a lesson to begin.
The lessons
- 01Turn a derivative into an updateUse a transparent quadratic to understand gradient descent before reading richer engines.Ready to learn12 min
- 02See why step size mattersDistinguish steady convergence, oscillation, and divergence in the same simple model.Ready to learn12 min
- 03Accept steps with a decrease testUse backtracking to reject an overshoot and choose a step that satisfies an explicit condition.Ready to learn12 min
- 04Understand penalties and tradeoffsSee how a regularizer changes the objective and its preferred solution.Ready to learn12 min
- 05Make an optimization run repeatableRecord enough state to distinguish a repeated experiment from a similar-looking animation.Ready to learn12 min
Progress is kept in this browser. How learning works