Extra parameter file

GADMA take an extra parameter file as input. However one probably do not need them. Nevertheless, if one is interested, extra_params_template with all options and their descriptions can be found in gadma/cli/ folder.

Related options

  • Options of parameter bounds:

    • Min N - minimum value (in genetic units) of population size (0.01).

    • Max N - maximum value (in genetic units) of population size (100).

    • Min T - minimum value (in genetic units) of epoch time (0).

    • Max T - maximum value (in genetic units) of epoch time (5).

    • Min M - minimum value (in genetic units) of migration rate (0).

    • Max M - maximum value (in genetic units) of migration rate (10).

  • Options of genetic algorithm:

    • Size of generation - number of demographic models on one iteration (generation) of the genetic algorithm (10).

    • Fractions - fractions of best, mutated and crossed models that are taken to the new generation (None).

    • N elitism - number of best models that are taken to the new generation (2).

    • P mutation - fraction of models that are mutants in new generation (0.56).

    • P crossover - fraction of models that are children in new generation (0.19).

    • P random - fraction of models that were randomly generated for new generation (0.13).

    • Mean mutation strength - initial value of mean number of parameters that are mutated in the chosen model during mutation (0.63).

    • Const for mutation strength - constant for ‘one-fifth’ rule to change mutation strength (1.02).

    • Mean mutation rate - initial value of mean rate of parameter change during mutation (0.45).

    • Const for mutation rate - constant for ‘one-fifth’ rule to change mutation rate (1.07).

    • Stuck generation number - genetic algorithm stops when there is no improvement during this number of iterations.

    • Eps - change of log-likelihood that is considered significant.

    • Random N_A - enables random generation of ancestral size value and scales other parameters to this value during random generation of model parameters (True).

  • Options of bayesian optimization:

    • Kernel - name of kernel for Gaussian process (Matern52).

    • Acquisition function - name of acquisition function (EI).

  • Options of global optimizations:

    • X_init - points for initial design.

    • Y_init - value of log-likelihood on this points.