A misclassified configuration for a ternary two-rule density classifier
Abstract
The ternary cellular automaton with Wolfram number 6478767664173 sends the configuration 210 of length 3 to 111, which it and the rule 7580606234490 both fix, although 210 contains a zero and has density 1/2. This refutes Conjecture 1 of H. Fukś and R. Procyk (arXiv:2002.08924), which asserts that this pair of rules classifies by density every finite configuration containing a zero.
Definition 1.1 (The local rule of a Wolfram number).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.wolfram (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
A ternary nearest-neighbour local rule is a map from {0,1,2}^3 to {0,1,2}. The rule with Wolfram number N has f(a,b,c) equal to the base-3 digit of N at the position 9a + 3b + c, that is, the integer part of N divided by 3^(9a+3b+c), reduced modulo 3; the paper indexes the coefficients as a_(9x_0+3x_1+x_2) = f(x_0,x_1,x_2). Here a, b, c lie in Fin 3, the division is division of natural numbers with remainder discarded, and the result is the element of Fin 3 with that value.
Definition 1.2 (The rule F).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.ruleF (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
F is the rule with Wolfram number 6478767664173; it preserves the sum of the cell values, and its restriction to the values 1 and 2 is elementary rule 184.
Definition 1.3 (The rule G).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.ruleG (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
G is the rule with Wolfram number 7580606234490.
Definition 1.4 (The global map on periodic configurations).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.step (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
A configuration of length L is a map x from ZMod L to Fin 3, so that indices are taken modulo L. The global map of a local rule f sends x to the configuration whose value at i is f(x(i - 1), x(i), x(i + 1)).
Definition 1.5 (The density).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.rho (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
For L different from 0 (NeZero L), the density of x is the rational number (1/2L) times the sum over i in ZMod L of the values x(i), read as natural numbers.
Definition 1.6 (Conjecture 1).
Formalization. D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.claim (✓ std3).
Citation. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
Conjecture 1 of the paper: for every length L (with NeZero L) and every configuration x of length L containing at least one zero, G^L F^L(x) is the constant configuration 0 if the density lies in [0, 2/3), the constant 1 if it lies in (2/3, 3/4), and the constant 2 if it lies in (3/4, 1). In the formula the exponent [L] denotes the L-fold iterate of the global map, and (i -> c) is the constant configuration c.
Theorem 1.7 (The configuration 210).
Proof. Machine-checked in Lean as D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.result (✓ std3). ∎
Resolves. Problems/fuks-2019-ternary-density-classification (refuted) by D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.result.
Source. Repository-derived.
Acknowledgement. Henryk Fukś; Roman Procyk (2019). Explorations of Ternary Cellular Automata and Ternary Density Classification Problems. DOI: 10.5506/APhysPolBSupp.12.75. URL: https://arxiv.org/abs/2002.08924v1.
Commentary.
Take L = 3 and x = (2, 1, 0), which contains a zero and has density 3/6 = 1/2, in [0, 2/3). Decoding the Wolfram number of F gives F(0,2,1) = F(2,1,0) = F(1,0,2) = 1, so the global map of F sends x to (1, 1, 1). Both rules send (1, 1, 1) to itself, since F(1,1,1) = G(1,1,1) = 1, so G^3 F^3(x) = (1, 1, 1) and not the constant 0 that the conjecture predicts. The iterates are evaluated by decide and the density by norm_num.
References
- Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.claim - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.result - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.rho - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.ruleF - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.ruleG - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.step - Truth anchor:
D5/S3/StatisticalMechanics/CellularAutomata/TernaryDensityClassificationRefutation.wolfram