The scaling factor for each run is incrementally increased until the structural median collapse (or until a defined failure limit state) is reached.
Five runs of four incrementally heated chains were run in parallel (temperature increment = 0.5).
Most stream classifiers are designed to process data incrementally, run in resource-aware environments, and react to concept drifts, i.e., unforeseen changes of the stream's underlying data distribution.
Bayesian analyses were initiated with random starting trees, run with four incrementally heated chains (Metropolis-coupled Markov chain Monte Carlo; [54] for 10 million generations, and sampled at intervals of 1000 generations. Two independent Bayesian analyses were run to avoid entrapment on local optima.
The MrBayes analysis of Dataset 1 implemented default priors and included two independent runs of four incrementally heated chains (t = 0.2) that were run for 6 million generations, while sampling trees from the current cold chain once every 100 generations.
In Bayesian analyses, two independent parallel runs of four incrementally heated Metropolis-coupled MCMCs (Monte Carlo Markov Chains) were run with trees sampled every 100 generations for 1,000,000 generations.
Two independent parallel runs of four incrementally heated Metropolis-coupled MCMCs (Monte Carlo Markov Chains) were run with trees sampled every 100 generations for 5 * 106 generations or more until to the average standard deviation of split frequency below 0.01.
In BI analyses conducted with MRBAYES, two independent runs of four incrementally heated Metropolis-coupled Markov chains Monte Carlo (MCMCMC) were launched and run for 5,000,000 generations.
Four separate MrBayes runs, each including four incrementally heated chains, were run for 20 million generations.
Bayesian tree topology and posterior probabilities (PP) were determined from two independent runs of four incrementally heated chains.
