search

Eawag-SIAM / SimulatedAnnealingABC.jl

Approximate Bayesian Computation algorithm based on simulated annealing
GNU General Public License v3.0
1 stars 0 forks source link

Automatically find `eps_init` and allow for multible summary statistics #1

Closed carloalbert closed 1 year ago

carloalbert commented 1 year ago

Get rid of eps_init:

  1. Sample n_particles from prior
  2. Calculate for each summery statistic: epsilon_init = mean(distance) (needs to be tested, maybe quantile better)
  3. compute acceptance prob: w = exp(-sum(distances ./ epsilon_inits))
  4. Build population using prior population and weights w
  5. Continue sampling until population is full
  6. Learn cdfs from prior sample for each summary stat
scheidan commented 1 year ago

closed in favor for #8