Download lean/OAI/Probability/LatticeSurface/Dynamics/ProjectionTail.lean from roseking/openAI-math-dataset: direct link, hf CLI and curl.
- Browser
- Download file 2.1 kB
-
https://huggingface.co/datasets/roseking/openAI-math-dataset/resolve/main/lean/OAI/Probability/LatticeSurface/Dynamics/ProjectionTail.lean
- Command line
-
hf download hf://datasets/roseking/openAI-math-dataset/lean/OAI/Probability/LatticeSurface/Dynamics/ProjectionTail.lean
-
curl -L -o ProjectionTail.lean https://huggingface.co/datasets/roseking/openAI-math-dataset/resolve/main/lean/OAI/Probability/LatticeSurface/Dynamics/ProjectionTail.lean
2.1 kB
| import Mathlib | |
| import OAI.Probability.LatticeSurface.Convexity.ElementaryKernels | |
| namespace OAI | |
| namespace ProjectionComparison | |
| open Set ClusterDisplacement | |
| variable {V E : Type*} [Fintype V] [Fintype E] | |
| theorem tail_projection_kernel (S : System V E) (b₀ b₁ v₀ v₁ h₀ h₁ : V → ℝ) (a : ℝ) | |
| (hp₀ : IsProjection S b₀ v₀ h₀) (hp₁ : IsProjection S b₁ v₁ h₁) | |
| (hb : ∀x∈S.pins,b₁ x-b₀ x≤a) : | |
| ∃ A : Kernel V,Firm A ∧ ∃ r : V → ℝ,(∀x,r x∈Icc (0:ℝ) 1) ∧ | |
| ∀x,max (h₁ x-h₀ x-a) 0=r x*A.apply (fun y => max (v₁ y-v₀ y-a) 0) x := by | |
| classical | |
| let u : V → ℝ := fun y => max (v₁ y-v₀ y-a) 0 | |
| let v : V → ℝ := fun y => v₀ y+u y | |
| let hne : (feasible S b₀).Nonempty := ⟨h₀,hp₀.1⟩ | |
| let k := project S b₀ hne v | |
| have hpk : IsProjection S b₀ v k := project_spec S b₀ hne v | |
| obtain ⟨A,hA,heq⟩:=projection_update_kernel S b₀ v₀ v h₀ k hp₀ hpk | |
| have he (x) : A.apply u x=k x-h₀ x := by | |
| simpa only [v,add_sub_cancel_left] using congrFun heq x | |
| have horder : ∀x,h₁ x≤k x+a := projection_order S b₁ b₀ v₁ v h₁ k a hp₁ hpk | |
| (fun x hx => by linarith [hb x hx]) | |
| (fun x hx => by dsimp [v,u];linarith [le_max_left (v₁ x-v₀ x-a) 0]) | |
| have hnn (x) : 0≤A.apply u x := Finset.sum_nonneg fun y _ => mul_nonneg (A.nonneg x y) (le_max_right _ _) | |
| have hle (x) : max (h₁ x-h₀ x-a) 0≤A.apply u x := by | |
| apply max_le | |
| · rw [he];linarith [horder x] | |
| · exact hnn x | |
| let r : V → ℝ := fun x => max (h₁ x-h₀ x-a) 0/A.apply u x | |
| refine ⟨A,hA,r,?_,?_⟩ | |
| · intro x | |
| constructor | |
| · exact div_nonneg (le_max_right _ _) (hnn x) | |
| · by_cases hz : A.apply u x=0 | |
| · simp [r,hz] | |
| · exact (div_le_one (lt_of_le_of_ne (hnn x) (Ne.symm hz))).mpr (hle x) | |
| · intro x | |
| by_cases hz : A.apply u x=0 | |
| · have hu : max (h₁ x-h₀ x-a) 0=0 := le_antisymm (by simpa only [hz] using hle x) (le_max_right _ _) | |
| simp only [hu,r,zero_div,zero_mul] | |
| · exact (div_mul_cancel₀ _ hz).symm | |
| end ProjectionComparison | |
| end OAI | |