Edit on GitHub
ParameterSpacePartitions.Options — Type
Options(;
    radius = .1, 
    bounds, 
    n_iters, 
    init_parms,
    parallel = false,
    adapt_radius! = adapt!,
    kwargs...
)

An object that holds configuration options for the MCMC sampler.

Fields

  • parallel=false: runs model on threads if true. A speed up is observed if the evaluation

time of the function is 1 ms or greater.

  • radius = .10: the initial radius length for each chain
  • bounds: a vector of tuples representing (lowerbound, upperbound) for each dimension in

the parameter space

  • x_range: the range of allowable values for each parameter
  • n_iters: number of iterations to perform
  • p_eval: the function that evalues the model and pattern functions
  • adapt_radius!=adapt!: a function in the form of func(chain, options; kwargs...) that adapts

the radius.

  • init_parms: a vector of starting points, such as [[.3,.4],[.3,.5]] in the 2 dimensional case.
  • n_dims: number of dimensions in parameter space
  • parm_names: a vector of symbols corresponding to parameter names. The default is [:p1,:p2,..:pn]
  • add_iters: the number of trials to run after merging chains with the same pattern located in the same region
source
ParameterSpacePartitions.adapt! — Method
adapt!(
    chain, 
    options; 
    t_rate = .20, 
    kwargs...
)

Iteratively adapts the radius to achieve a target acceptance rate. The radius is adjusted according to the following factor c:

c = exp(λ * d_rate)

where λ is the adaption rate, and d_rate is the difference between the acceptance rate and target acceptance rate.

Arguments

  • chain: a chain for exploring the parameter space
  • options: a set of options for configuring the algorithm

Keyword Arguments

  • t_rate = .20: target acceptance rate
  • kwargs...: keyword arguments that are not processed
source
ParameterSpacePartitions.estimate_volume — Method
estimate_volume(
    model, 
    p_fun, 
    points, 
    target, 
    bounds,  
    args...;
    n_sim = 10_000,
    kwargs...
)

Estimate volume of region with an eillipsoid and hit or miss bias adjustment.

Arguments

  • model: a model function that returns predictions given a vector of parameters
  • p_fun: a function that that evaluates the qualitative data pattern
  • points: a p x n matrix of sampled points where p is the number of parameters and n is the number of samples
  • target: the target pattern associated with the points
  • bounds: a vector of tuples representing (lowerbound, upperbound) for each dimension in
  • args...: arguments passed to model and p_fun

Keywords

  • n_sim=10_000: the number of samples for hit or miss bias adjustment
  • kwargs...: additional keyword arguments passed to model or p_fun
source
ParameterSpacePartitions.find_partitions — Method
find_partitions(model, p_fun, options, args...; show_timer=false, kwargs...)

Performs parameter space partitioning.

Arguments

  • model: a model function that returns predictions given a vector of parameters
  • p_fun: a function that that evaluates the qualitative data pattern
  • options: a set of options for configuring the algorithm
  • args...: arguments passed to model and p_fun

Keywords

  • show_timer=false: displays timer and progress bar if true
  • kwargs...: keyword arguments passed to model and p_fun
source