BBO-LP - example

[theo20]

Hello. I want to maximise my function goldstein, using batch mode and Local Penalisation. But I get an error. Could you please tell me why this might not working? I obtained the results but I am not sure if these are correct.

1. I still don't know from where we get the optimal value at each iteration. If I specify my batch size =4, then at each iteratation I get four new points. From where I choose these 4 points?
2. What is the optimisation restarts and what the acquisition_weight?

Here is the code that I am using and the error that I obtain.
Thank you in advance for your help.

#define my objective function
def goldstein(D):
....
myfun=lambda var: goldstein(var);

#domain
domain =[{'name': 'var_1', 'type': 'continuous', 'domain': (-2,2)},
{'name': 'var_2', 'type': 'continuous', 'domain': (-2,2)}]

#initial design
space = GPyOpt.Design_space(domain)
data_init = 10
X = initial_design('latin',space,data_init)

#Evaluate my function at X
Y = myfun(X)

#Batch Size & Iterations
batch_size = 4
max_iter = 10
num_cores = 4

#Optimisation using BBO-LP
myBopt = GPyOpt.methods.BayesianOptimization(f=goldstein, # function to optimize
domain=domain,
X=X,
Y=Y,
acquisition_type='EI',
exact_feval = True,
normalize_Y = True,
evaluator_type = 'local_penalization',
batch_size = batch_size,
num_cores = num_cores,
optimize_restarts = 10,
acquisition_weight = 2,
de_duplication = True)

myBopt.run_optimization(max_iter)

EOFError
Error in parallel computation. Fall back to single process!