Download lean/OAI/NumberTheory/TwoPoint/Circuits/CircuitLogarithmicBudget.lean from roseking/openAI-math-dataset: direct link, hf CLI and curl.
- Browser
- Download file 2.84 kB
-
https://huggingface.co/datasets/roseking/openAI-math-dataset/resolve/main/lean/OAI/NumberTheory/TwoPoint/Circuits/CircuitLogarithmicBudget.lean
- Command line
-
hf download hf://datasets/roseking/openAI-math-dataset/lean/OAI/NumberTheory/TwoPoint/Circuits/CircuitLogarithmicBudget.lean
-
curl -L -o CircuitLogarithmicBudget.lean https://huggingface.co/datasets/roseking/openAI-math-dataset/resolve/main/lean/OAI/NumberTheory/TwoPoint/Circuits/CircuitLogarithmicBudget.lean
2.84 kB
| import OAI.NumberTheory.TwoPoint.Circuits.CircuitPolynomialGrowth | |
| namespace OAI | |
| /-! Convert a dyadic accuracy index to the natural logarithm used in the | |
| published fixed-depth bounded-independence input. -/ | |
| namespace TwoPointCorrelations | |
| lemma dyadic_index_of_ratio (m : ℕ) (hm : 1 ≤ m) (ε : ℝ) | |
| (hε : 0 < ε) (hεhalf : ε ≤ 1 / 2) : | |
| ∃ j : ℕ, m ≤ 2 ^ j ∧ (1 / 2 : ℝ) ^ j ≤ ε ∧ | |
| (j : ℝ) + 1 ≤ 6 * Real.log ((m : ℝ) / ε) := by | |
| let L := Real.log ((m : ℝ) / ε) | |
| have hmreal : (1 : ℝ) ≤ m := by exact_mod_cast hm | |
| have hratio : (2 : ℝ) ≤ (m : ℝ) / ε := by | |
| apply (le_div_iff₀ hε).mpr | |
| linarith | |
| have hlogtwo : (1 / 2 : ℝ) ≤ Real.log 2 := by | |
| have h := Real.one_sub_inv_le_log_of_pos (by norm_num : (0 : ℝ) < 2) | |
| norm_num at h ⊢ | |
| exact h | |
| have hL : 1 / 2 ≤ L := hlogtwo.trans (Real.log_le_log (by norm_num) hratio) | |
| let j := ⌈2 * L⌉₊ | |
| have hjlo : 2 * L ≤ (j : ℝ) := Nat.le_ceil _ | |
| have hjhi : (j : ℝ) < 2 * L + 1 := Nat.ceil_lt_add_one (by linarith) | |
| have hlogbound : L ≤ (j : ℝ) * Real.log 2 := by | |
| nlinarith [show (0 : ℝ) ≤ j from Nat.cast_nonneg j] | |
| have hx : (m : ℝ) / ε ≤ (2 : ℝ) ^ j := by | |
| apply (Real.log_le_log_iff (by positivity) (by positivity)).mp | |
| simpa only [Real.log_pow] using hlogbound | |
| have hmul : (m : ℝ) ≤ (2 : ℝ) ^ j * ε := (div_le_iff₀ hε).mp hx | |
| have hmle : (m : ℝ) ≤ (2 : ℝ) ^ j := by nlinarith [show (0 : ℝ) < 2 ^ j by positivity] | |
| have heps : (1 / 2 : ℝ) ^ j ≤ ε := by | |
| have hh : (1 : ℝ) / (2 : ℝ) ^ j ≤ ε := | |
| (div_le_iff₀ (by positivity)).mpr (by nlinarith) | |
| simpa only [div_pow, one_pow] using hh | |
| exact ⟨j, by exact_mod_cast hmle, heps, by linarith⟩ | |
| theorem bravermanDegree_logarithmic_budget : | |
| ∃ K C : ℕ, 0 < K ∧ 0 < C ∧ ∀ m : ℕ, 1 ≤ m → | |
| ∀ ε : ℝ, 0 < ε → ε ≤ 1 / 2 → ∀ t : ℕ, | |
| (K : ℝ) * (Real.log ((m : ℝ) / ε)) ^ C ≤ (t : ℝ) → | |
| ∃ j : ℕ, m ≤ 2 ^ j ∧ (1 / 2 : ℝ) ^ j ≤ ε ∧ bravermanDegree j ≤ t := by | |
| obtain ⟨K, C, hK, hC, hbound⟩ := bravermanDegree_polynomial_positive | |
| refine ⟨K * 6 ^ C, C, by positivity, hC, ?_⟩ | |
| intro m hm ε hε hεhalf t ht | |
| obtain ⟨j, hmj, hjε, hjlog⟩ := dyadic_index_of_ratio m hm ε hε hεhalf | |
| refine ⟨j, hmj, hjε, ?_⟩ | |
| have hj : (bravermanDegree j : ℝ) ≤ (K : ℝ) * ((j : ℝ) + 1) ^ C := by | |
| exact_mod_cast hbound j | |
| have hdeg : (bravermanDegree j : ℝ) ≤ (t : ℝ) := by | |
| calc | |
| _ ≤ _ := hj | |
| _ ≤ (K : ℝ) * (6 * Real.log ((m : ℝ) / ε)) ^ C := by gcongr | |
| _ = ((K * 6 ^ C : ℕ) : ℝ) * (Real.log ((m : ℝ) / ε)) ^ C := by | |
| push_cast | |
| rw [mul_pow] | |
| ring | |
| _ ≤ _ := ht | |
| exact_mod_cast hdeg | |
| end TwoPointCorrelations | |
| end OAI | |