Consensus and fixation #
Configurations are maps s : V → C (a colour, opinion or type per vertex). A consensus is a
constant configuration, and nonConsensus is the indicator of not being one: the survival
indicator for which the absorption theorem (Kernel.finite_absorption) is applied in the model
packages (Voter.survival, Moran.unfixed, Undecided.notMono).
Kernel.nonConsensus_step: if consensus configurations are absorbing,K nonConsensus ≤ nonConsensus, the absorption hypothesis offinite_absorption.Kernel.iterate_nonConsensus_bool: for two colours, finite-time non-consensus is the event of being neither all-truenor all-false.Kernel.iSup_consensus_of_harmonic: the gambler's ruin formula, fixation in all-truefrom a harmonic observable (the two-consensus case ofiSup_event_of_invariant), and its normalized formKernel.iSup_consensus_of_invariant.
If every consensus configuration is absorbing, the non-consensus probability does not increase in one step.
For two colours, finite-time non-consensus is an event.
Gambler's ruin for two consensus configurations. If φ is harmonic for K and takes
different values on all-true and all-false, all-true is absorbing, and the non-consensus
probability tends to zero, then the probability of fixation in all-true from s is
(φ(s) − φ(all-false)) / (φ(all-true) − φ(all-false)). For instance, a harmonic q^k, with k
the number of true vertices out of n, gives (1 - q^k) / (1 - q^n), and a harmonic k gives
k / n. This is iSup_event_of_invariant for the two consensus configurations.
Fixation from an invariant. If ψ is harmonic for K, equals 1 on all-true and 0
on all-false, all-true is absorbing, and the non-consensus probability tends to zero, then
ψ is the probability of fixation in all-true. This is the normalized case of
iSup_consensus_of_harmonic.