DEMCMC

Purpose

DEMCMC uses Differential Evolution proposals across chains.

Full Sampling Section Keys

• Sampling.Method (required): must be DEMCMC.
• Sampling.Variables (required, array):
• name (required)
• description (required)
• distribution.type (required)
• distribution.parameters (required)
• runtime-safe parameter sets:
  • Flat: min, max
  • Log: min, max
  • Normal: mean, stddev
  • Log-Normal: mean, stddev
  • Logit: location, scale
• Sampling.LogLikelihood (required): {name, expression} list.
• Sampling.selection (optional, string)
• Sampling.Bounds:
• num_chains (required, integer)
• num_iters (required, integer)
• proposal_scale (required, number or array)
• de_gamma (optional, number, default 0.0)
• de_noise (optional, number, default 1e-3)
• de_crossover (optional, number, default 1.0)

Full Skeleton

Sampling:
  Method: "DEMCMC"
  Variables:
    - name: p1
      description: parameter 1
      distribution:
        type: Flat
        parameters:
          min: -5
          max: 5
    - name: p2
      description: parameter 2
      distribution:
        type: Flat
        parameters:
          min: -5
          max: 5
  LogLikelihood:
    - name: L_total
      expression: "L1 + L2"
  Bounds:
    num_chains: 10
    num_iters: 12000
    proposal_scale: [0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10]
    de_gamma: 0.7
    de_noise: 0.001
    de_crossover: 0.9

Example

Sampling:
  Method: "DEMCMC"
  Variables:
    - name: m0
      description: scalar mass
      distribution:
        type: Flat
        parameters:
          min: 100
          max: 3000
    - name: m12
      description: gaugino mass
      distribution:
        type: Flat
        parameters:
          min: 100
          max: 3000
  LogLikelihood:
    - name: L_total
      expression: "L_higgs + L_dm"
  Bounds:
    num_chains: 10
    num_iters: 12000
    proposal_scale: [0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10, 0.10]
    de_gamma: 0.7
    de_noise: 0.001
    de_crossover: 0.9