/BOMI

Bayesian Optimization with Missing Inputs

Primary LanguagePython

Bayesian Optimization with Missing Inputs

The implementation of BOMI in the paper 'Bayesian Optimization with Missing Inputs', ECMLPKDD2020.

Prerequisites

  • Python 3.6
  • Numpy 1.18
  • Scipy 1.3.1
  • Scikit-learn 0.21.2
  • Torch 1.3.1 (CUDA v9.2)
  • Gpytorch 1.0.1
  • Missingpy 0.2.0
  • Pandas 0.25.3

(Optional)

  • pip 19.3.1
  • pillow 5.4.1

Instruction

  • Execute an experiment with the command: ..\python runExperiment <opt_method> <obj_function> <num_of_GPs> <alpha_param> <miss_rate> <miss_noise>

Example:

python runExperiemnt BOMI Eggholder2d 5 1e2 0.25 0.05

See 'runExperiment.py' for more details and see 'ndfunction.py' for how to define a new objective function.

Reference

License

Apache 2.0