From: Learning versus optimal intervention in random Boolean networks
Parameter | Value | Brief Description |
---|---|---|
R | 1400 | Rule population size |
γ | 0.71 | Discount rate |
θmna | N+1 | Min. number of actions in match set |
P# | 0.33 | Probability of hash |
pI | 0.0 | Initial payoff |
εI | 0.0 | Initial error |
FI | 0.0 | Initial fitness |
ε0 | 10 | Error threshold |
θga | 5.0 | Genetic algorithm frequency |
θdel | 20.0 | Deletion threshold |
β | 0.2 | Affects update of p, ε, and action set size for classifiers |
α | 0.1 | Affects fitness updates |
ν | 5.0 | Affects fitness updates |
χ | 0.8 | Likelihood of GA crossover operation |
μ | 0.05 | Likelihood of GA mutation operation |
δ | 0.1 | Mods. effect of fitness on the ‘deletion vote’ of a classifier |
θsub | 30.0 | Subsumption threshold |
pexplr | 0.5 | Likelihood of exploring |
doAsSubsumpt. | true | Perform subsumption in the action set? |
doGaSubsumpt. | true | Perform subsumption in the GA? |