Akari

AbbreviationAkari boards#✓fully proved

An Akari instance is a board given as a list of rows, each row a list of natural numbers: the entry \(0\) is an empty cell, \(1\) is a wall, and \(2 + k\) is a wall numbered \(k\). The rows need not have the same length; that is part of well-formedness.

abbrev AkariInstance := List (List ℕ)
Used by 3

DefinitionIllumination on an Akari board#✓fully proved

Let \(B\) be a board, a list of rows of natural numbers, and write \(B_{r,c}\) for the entry in row \(r\) and column \(c\), read as \(1\) when there is no such entry. A bulb at \((b_r, b_c)\) illuminates the cell \((r, c)\) when either \(b_r = r\) and \(B_{r,c'} = 0\) for every column \(c'\) strictly between \(b_c\) and \(c\), or \(b_c = c\) and \(B_{r',c} = 0\) for every row \(r'\) strictly between \(b_r\) and \(r\). In particular, a bulb illuminates its own cell.

def Illuminates (board : List (List ℕ)) (br bc r c : ℕ) : Prop :=
  (br = r ∧ ∀ cc, (min bc c < cc ∧ cc < max bc c) →
    (board.getD r []).getD cc 1 = 0) ∨
  (bc = c ∧ ∀ rr, (min br r < rr ∧ rr < max br r) →
    (board.getD rr []).getD c 1 = 0)
Used by 1

DefinitionWell-formed Akari boards#✓fully proved

An Akari board is well formed when every entry is at most \(6\) and all its rows have the same length.

def WellFormed (i : AkariInstance) : Prop :=
  ∀ row ∈ i, ∀ x ∈ row, x ≤ 6 ∧ ∀ row ∈ i, row.length = (i.getD 0 []).length
Uses 1 · Used by 1

DefinitionFeasible Akari boards#✓fully proved

Let \(B\) be an Akari board with \(m\) rows, let \(n\) be the length of its first row (\(n = 0\) when there are no rows), and write \(B_{r,c}\) for the entry in row \(r\) and column \(c\), read as \(1\) when there is no such entry. The board \(B\) is feasible when there is a finite set \(S\) of pairs of natural numbers (the bulbs) such that every bulb \((r, c) \in S\) has \(r < m\), \(c < n\) and \(B_{r,c} = 0\); every cell \((r, c)\) with \(r < m\), \(c < n\) and \(B_{r,c} = 0\) is illuminated by some bulb in \(S\); no bulb in \(S\) illuminates a different bulb in \(S\); and for every \((r, c)\) with \(r < m\), \(c < n\) and \(B_{r,c} \ge 2\), the number of bulbs \((r', c') \in S\) with \(|r - r'| + |c - c'| = 1\) equals \(B_{r,c} - 2\).

def AkariFeasible
    (i : AkariInstance) : Prop :=
  ∃ bulbs : List (ℕ × ℕ), bulbs.Nodup ∧
    (∀ p ∈ bulbs, p.1 < i.length ∧ p.2 < (i.getD 0 []).length ∧
      (i.getD p.1 []).getD p.2 1 = 0) ∧
    (∀ r < i.length, ∀ c < (i.getD 0 []).length, (i.getD r []).getD c 1 = 0 →
      ∃ b ∈ bulbs, Illuminates i b.1 b.2 r c) ∧
    (∀ b ∈ bulbs, ∀ b' ∈ bulbs, b ≠ b' → ¬ Illuminates i b.1 b.2 b'.1 b'.2) ∧
    ∀ r < i.length, ∀ c < (i.getD 0 []).length, 2 ≤ (i.getD r []).getD c 1 →
      ((bulbs.filter (fun b => (b.1 == r + 1 && b.2 == c) || (b.1 + 1 == r && b.2 == c) ||
        (b.1 == r && b.2 == c + 1) || (b.1 == r && b.2 + 1 == c))).length =
        (i.getD r []).getD c 1 - 2)
Uses 2 · Used by 1

DefinitionThe Akari decision problem#✓fully proved

The Akari (Light Up) decision problem accepts a list of booleans \(x\) exactly when \(x\) decodes, under the canonical bitstring encoding of lists of lists of natural numbers, to an Akari board that is well formed and feasible; it rejects every list that does not decode.

noncomputable def akariDec :
    MetaMathlibExt.ComplexityTheory.DecisionProblem := fun x =>
  match MetaMathlibExt.BitstringEncoding.bitDecode
      (α := AkariInstance) x with
  | some i => decide (WellFormed i ∧ AkariFeasible i)
  | none => false
Uses 12 · Used by 1

TheoremAKARI (LIGHT UP) is NP-complete#●ready to prove

The Akari decision problem belongs to NP, and every decision problem in NP reduces to it by a polynomial-time many-one reduction.

public theorem_wanted akari_npComplete :
    akariDec ∈ MetaMathlibExt.ComplexityTheory.NP ∧
      ∀ L : MetaMathlibExt.ComplexityTheory.DecisionProblem,
        L ∈ MetaMathlibExt.ComplexityTheory.NP →
          Cslib.CookLevin.PolyManyOneRed L akariDec
Sources
  • Brandon McPhail, "Light Up is NP-complete", unpublished manuscript, 2005.
Uses 5
5 fully proved1 ready to prove

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.