CODEWITHSUNDEEP
roptimizationmlr3black-box
JacquieS
JacquieS

Reputation: 43

MLR3MBO - function with inputs not to optimise

When defining the objective function to be optimised in MLR3MBO, is it possible for it to have inputs NOT to be optimised?

Longer description: Data: humans playing a computer task, so a series of y values (the buttons they pressed) and at the same time x values (what was shown on the screen). I have a model (function) of their cognition (human decision-making model), i.e. that describes how y is created from x, given parameters p. I want to optimise p, i.e. find the 'right' parameters to describe each single participant. Each participant has thus different x, y and p.

I have tried to capture this by giving list xs to the function to optimise with

xs=list(x1=x1,x2=x2... ,p1=p1,p2=p2)
domain=paradox::ps(p1=p_dbl(lower = -10, upper = 10),....)

So not including x1,x2... in the domain definition. But when I then run the code (see below), it does not recognise x1,x2...

codomain=ps(y=p_dbl(tags='minimize'))
  objective= ObjectiveRFun$new(
    fun=myModel,
    domain=domain,
    codomain=codomain)
  instance = OptimInstanceSingleCrit$new(
    objective=objective,
    search_space=domain,
    terminator=trm('evals',n_evals =60)) # maybe change this?
  # Gaussian Process, EI, DIRECT
  surrogate = srlrn(lrn("regr.km",
                        covtype = "matern3_2",
                        optim.method = "gen",
                        nugget.stability = 10^-8, control = list(trace = FALSE)))
  acq_function = acqf("ei")
  acq_optimizer = acqo(opt("nloptr", algorithm = "NLOPT_GN_DIRECT_L"),
                       terminator = trm("stagnation", threshold = 1e-8))
  optimizer = opt("mbo",
                  loop_function = bayesopt_ego,
                  surrogate = surrogate,
                  acq_function = acq_function,
                  acq_optimizer = acq_optimizer)

  set.seed(2906)
  start.time=Sys.time()
  optimizer$optimize(instance)

The reason I thought of using MLR3MBO in the first place is that the function takes very long to compute on each optimization loop (it's got a for loop with many steps).

Upvotes: 0

Views: 26

Answers (0)

Related Questions