cma.fitness_models.SurrogatePopulation

surrogate f-values for a population.

See also __call__ method.

What is new:

• the model is built with >=3 evaluations (compared to LS-CMA and lmm-CMA)
• the model is linear at first, then diagonal, then full
• the model uses the fitness ranks as weights for weighted regression.

>>> import cma
>>> import cma.fitness_models as fm
>>> from cma.fitness_transformations import Function as FFun  # adds evaluations attribute
>>> # fm.Logger, Logger = fm.LoggerDummy, fm.Logger
>>> surrogate = fm.SurrogatePopulation(cma.ff.elli)

Example using the ask-and-tell interface:

>>> for ifun, fun in enumerate([cma.ff.elli, cma.ff.sectorsphere]):
...     for irun in range(7):
...         assert irun < 4  # measure best of four
...         es = cma.CMAEvolutionStrategy(5 * [1], 2.2,
...                     {'CMA_injections_threshold_keep_len': 1,
...                         'ftarget':1e-9, 'verbose': -9, 'seed':3 + irun})
...         fitfun = FFun(fun)
...         surrogate = fm.SurrogatePopulation(fitfun)
...         while not es.stop():
...             X = es.ask()
...             es.tell(X, surrogate(X))  # surrogate evaluation
...             es.inject([surrogate.model.xopt])
...             # es.disp(); es.logger.add()  # ineffective with verbose=-9
...         # print(irun, fitfun.evaluations)  # was: (sig=2.2) 12 161, 18 131, 18 150, 18 82, 15 59, 15 87, 15 132, 18 83, 18 55, 18 68
...         assert 'ftarget' in es.stop()
...         if ifun == 0:
...             if fitfun.evaluations < 20:
...                 break
...         if ifun == 1:
...             if fitfun.evaluations < 90:
...                 break

Example using the parallel_objective interface to cma.fmin:

>>> for fitfun, evals in [[FFun(cma.ff.elli), 22], [FFun(cma.ff.ellirot), 40]]:
...     surrogate = fm.SurrogatePopulation(fitfun)
...     inject_xopt = fm.ModelInjectionCallback(surrogate.model)  # must use the same model
...     xopt, es = cma.fmin2(None, 5 * [1], 2.2,
...                      {'CMA_injections_threshold_keep_len': 1,
...                       'ftarget':1e-12, 'verbose': -9},
...                      parallel_objective=surrogate,
...                      callback=inject_xopt)
...     # print(fitfun.evaluations)
...     assert fitfun.evaluations == es.result.evaluations
...     assert es.result[1] < 1e-12
...     assert es.result[2] < evals
>>> # fm.Logger = Logger