Milestone map
Milestone map
2 milestones
Do the problem sets and write them up
5–10 weeks (8–10 hrs/week)
Take Optimization theory for free and do the book's exercises on convex sets, convex functions, duality and the KKT conditions. Write every solution up in LaTeX and keep the source and the PDF in one public repo as you go.
Proof required
Link your public repo with the LaTeX source and the PDF of the book's exercises on convex sets, convex functions, duality and the KKT conditions, and show your self-check (Check each answer with a solver).
What gets checked
- Every convexity proof is complete.
- Dual problems are derived step by step.
- Solver answers agree with hand answers.
Common mistakes
- Copying posted solutions: verifiers read the reasoning and the history.
- A proof that skips the hard step.
- Asking for review before the self-check.
Powstik Guide
A public repo with your LaTeX write-ups of the book's exercises on convex sets, convex functions, duality and the KKT conditions.
Steps
- Make a public repo for this course.
- Do one problem set at a time; typeset it in LaTeX and commit.
- Self-check: Solve each problem by hand, then check the answer with CVXPY (free) on the same problem.
- Fix what the self-check shows, and say what you fixed.
- Ask three people who know the topic to check it.
Template
README Course: Optimization theory (Convex Optimization by Boyd and Vandenberghe (free book and Stanford EE364a materials)) What is here (problem sets, PDF): Self-check result: The hardest proof and why:
What gets sent back
- Answers with no reasoning.
- Photos of handwriting instead of typeset work.
- Self-check not shown.
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Every convexity proof is complete.
- Dual problems are derived step by step.
- Solver answers agree with hand answers.
- Pick two proofs and check every step: each argument must be correct and complete, not just present.
- Ask about one step; the answer should be clear.
You'll sign in first, then come straight back here.
Solve a real problem by convex optimisation
1–3 weeks (5–8 hrs/week)
Use what you learned. Make a modelling notebook that states a real problem as a convex program (a diet, a portfolio, a fit), proves it is convex, derives its dual and solves both. Typeset it in LaTeX (or a notebook with LaTeX maths) in your public repo.
Proof required
Link a modelling notebook that states a real problem as a convex program (a diet, a portfolio, a fit), proves it is convex, derives its dual and solves both, in your public repo.
What gets checked
- Convexity is proved.
- The dual is derived correctly and strong duality holds (shown).
- The solution makes sense for the real problem.
Common mistakes
- A gap in the main argument.
- A result stated but not proved.
Powstik Guide
A modelling notebook that states a real problem as a convex program (a diet, a portfolio, a fit), proves it is convex, derives its dual and solves both.
Steps
- Write one line: what you will prove or show.
- Draft it, then read it cold the next day.
- Ask someone to read it and mark the first step they doubt.
- Fix it and typeset the final version.
- Ask three people to check it.
Template
Title: What this shows: Main result (statement): Proof: What I would look at next:
What gets sent back
- A main step left unproved.
- Not clearly the person's own work.
- Claims more than it proves.
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Convexity is proved.
- The dual is derived correctly and strong duality holds (shown).
- The solution makes sense for the real problem.
- Every argument must be correct and complete; mark the first gap you find.
- Check that the work is the person's own.
You'll sign in first, then come straight back here.