Lemma 4
Lax496464.Lemma4 · concepts/Lax496464/Lemma4.lean · lax-496464
No public endorsements yet.
Loading review…
Sign in with ORCIDNatural Language Statement
Theorem
After the greedy has considered the first jobs, the set it holds is feasible, uses only those jobs, and dominates every feasible set of jobs among them. Consequently the set it holds at the end is a feasible set of largest possible cardinality, which is the correctness of the algorithm of Section 6.1.
Concept map
Evidence
In the paper
- page 6 of this submission's paper
Lean source view on GitHub
| 1 | import Lax496464.Greedy |
| 2 | |
| 3 | /-! |
| 4 | --- |
| 5 | title: Lemma 4 |
| 6 | type: theorem |
| 7 | --- |
| 8 | After the greedy has considered the first jobs, the set it holds is feasible, uses |
| 9 | only those jobs, and dominates every feasible set of jobs among them. Consequently the |
| 10 | set it holds at the end is a feasible set of largest possible cardinality, which is the |
| 11 | correctness of the algorithm of Section 6.1. |
| 12 | |
| 13 | # Formalization Notes |
| 14 | |
| 15 | The statement is about every run of the relation, since the rule leaves choices open, and |
| 16 | it is proved by induction along the run with the domination property as the invariant. |
| 17 | |
| 18 | The order is the counting one, which is stronger than what the paper's disjunction gives |
| 19 | where the cardinalities differ — see the definition for why the paper's will not carry |
| 20 | this induction. The conclusion of the algorithm's correctness is unaffected: an order |
| 21 | that dominates in this sense also dominates in cardinality. |
| 22 | |
| 23 | The run is presented as a sequence of sets indexed by how many jobs have been considered, |
| 24 | with a first member that is empty and a step at each job. Nothing requires the sequence |
| 25 | to be computed by anything in particular; what is claimed is that any sequence satisfying |
| 26 | the two conditions ends at an optimum. |
| 27 | |
| 28 | Equal preprocessing times and positive processing times are hypotheses, as in the |
| 29 | paper. Weights play no role: this is the unweighted case, and the conclusion is about |
| 30 | cardinality. |
| 31 | -/ |
| 32 | |
| 33 | namespace Lax496464.Lemma4 |
| 34 | |
| 35 | open Lax496464.FlowShop Lax496464.FlowShop.Instance |
| 36 | open Lax496464.EstOrder Lax496464.Greedy |
| 37 | |
| 38 | variable (I : Instance) |
| 39 | |
| 40 | /-- **Lemma 4.** Along any greedy run, the set held after `k` jobs is a feasible subset of |
| 41 | the first `k` jobs dominating every such feasible set. -/ |
| 42 | axiom greedy_dominating (hest : EstOrdered I) {p : ℕ} (hp : ∀ i : I.Job, I.p i = p) |
| 43 | (hq : ∀ i : I.Job, 0 < I.q i) |
| 44 | (S : ℕ → Finset I.Job) (h0 : S 0 = ∅) |
| 45 | (hrun : ∀ k, ∀ hk : k < I.jobs, Step I p (S k) (S (k + 1)) ⟨k, hk⟩) : |
| 46 | ∀ k ≤ I.jobs, |
| 47 | Feasible I (S k) ∧ S k ⊆ firstJobs I k ∧ |
| 48 | ∀ B : Finset I.Job, B ⊆ firstJobs I k → Feasible I B → SDom I (S k) B |
| 49 | |
| 50 | /-- **The greedy is optimal.** Its final set is a feasible set of largest cardinality. -/ |
| 51 | axiom greedy_card_max (hest : EstOrdered I) {p : ℕ} (hp : ∀ i : I.Job, I.p i = p) |
| 52 | (hq : ∀ i : I.Job, 0 < I.q i) |
| 53 | (S : ℕ → Finset I.Job) (h0 : S 0 = ∅) |
| 54 | (hrun : ∀ k, ∀ hk : k < I.jobs, Step I p (S k) (S (k + 1)) ⟨k, hk⟩) : |
| 55 | Feasible I (S I.jobs) ∧ |
| 56 | ∀ B : Finset I.Job, Feasible I B → B.card ≤ (S I.jobs).card |
| 57 | |
| 58 | end Lax496464.Lemma4 |
| 59 |
Formalization Notes
The statement is about every run of the relation, since the rule leaves choices open, and it is proved by induction along the run with the domination property as the invariant.
The order is the counting one, which is stronger than what the paper's disjunction gives where the cardinalities differ — see the definition for why the paper's will not carry this induction. The conclusion of the algorithm's correctness is unaffected: an order that dominates in this sense also dominates in cardinality.
The run is presented as a sequence of sets indexed by how many jobs have been considered, with a first member that is empty and a step at each job. Nothing requires the sequence to be computed by anything in particular; what is claimed is that any sequence satisfying the two conditions ends at an optimum.
Equal preprocessing times and positive processing times are hypotheses, as in the paper. Weights play no role: this is the unweighted case, and the conclusion is about cardinality.
Discussion
Ask a question or add context. Endorsements and structured flags are kept in the review panel above.
0 comments