active_sample#
- pyvbmc.vbmc.active_sample(gp: GP, sample_count: int, optim_state: dict, function_logger: FunctionLogger, iteration_history: IterationHistory, vp: VariationalPosterior, options: Options)[source]#
Actively sample points iteratively based on acquisition function.
- Parameters:
- gpgpyreg.GP or None
The GaussianProcess from the VBMC instance this function is called from.
Noneruns the initial design instead of the acquisition: the starting points and, where they fall short of sample_count, points drawn in the plausible box.- sample_countint
The number of samples.
- optim_statedict
The optim_state from the VBMC instance this function is called from.
- function_loggerFunctionLogger
The FunctionLogger from the VBMC instance this function is called from.
- iteration_historyIterationHistory
The IterationHistory from the VBMC instance this function is called from.
- vpVariationalPosterior
The VariationalPosterior from the VBMC instance this function is called from.
- optionsOptions
Options from the VBMC instance this function is called from.
- Returns:
- function_loggerFunctionLogger
The updated FunctionLogger.
- optim_statedict
The updated optim_state.
- vpVariationalPosterior
The updated VP.
- gpgpyreg.GP or None
The updated GP;
Noneafter the initial design.
Notes
Random draws (initial design, acquisition choice, search points, CMA-ES) use
vp.rng.