Simplex Ladder¶
Who This Is For¶
Use this ladder when the optimization story is already continuous and linear, and you now need one contest-time LP solver instead of one more proof-only lens.
Warm-Up¶
- normalize every constraint into
Ax <= b - tell continuous LP apart from integer programming
Core¶
- variable-per-product or variable-per-plan modeling
- trust the three status outcomes:
optimal,infeasible,unbounded
Repo Anchors¶
Stretch¶
- read one LP-heavy official statement and decide whether the numeric solver is actually the right route
- explain why one flow-style task should not be solved with generic simplex even if it can be written as an LP
Compare Points¶
- direct continuous LP solver route -> Cheese, If You Please
- benchmark / certificate lens without a solver -> Optimization And Duality
- network-specialized optimization instead of generic LP -> MINCOST
- exact-count relaxation on a DP family instead of a dense tableau -> Red and Blue Lamps
This ladder is intentionally sparse. The point is not to collect many simplex tasks. The point is to know when one LP solver is the cleanest honest route.
Exit Criteria¶
You are ready to move on when you can:
- translate a blend / diet / resource-allocation statement into
maximize c^T x, Ax <= b, x >= 0 - explain the meanings of
optimal,infeasible, andunbounded - reject simplex quickly when the task is really flow, DP, or integer programming