numpy truth value error

[himat]

I followed the example on the gpyopt website, and immediately got an error.

$ python
Python 2.7.10 (default, Feb  6 2017, 23:53:20) 
[GCC 4.2.1 Compatible Apple LLVM 8.0.0 (clang-800.0.34)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> # --- Load GPyOpt
... from GPyOpt.methods import BayesianOptimization
>>> import numpy as np
>>> 
>>> def f(x): return (6*x-2)**2*np.sin(12*x-4)
...
>>> bounds = [(0,1)]
>>> myBopt = BayesianOptimization(f=f, bounds=bounds)
>>> myBopt.run_optimization(max_iter=15)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/himakee/Library/Python/2.7/lib/python/site-packages/GPyOpt/methods/bayesian_optimization.py", line 458, in run_optimization
    super(BayesianOptimization, self).run_optimization(max_iter = max_iter, max_time = max_time,  eps = eps, verbosity=verbosity, save_models_parameters = save_models_parameters, report_file = report_file, evaluations_file= evaluations_file, models_file=models_file)
  File "/Users/himakee/Library/Python/2.7/lib/python/site-packages/GPyOpt/core/bo.py", line 103, in run_optimization
    self._update_model()
  File "/Users/himakee/Library/Python/2.7/lib/python/site-packages/GPyOpt/core/bo.py", line 198, in _update_model
    self._save_model_parameter_values()
  File "/Users/himakee/Library/Python/2.7/lib/python/site-packages/GPyOpt/core/bo.py", line 209, in _save_model_parameter_values
    if self.model_parameters_iterations == None:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
>>>

[arutaku]

@himat I got the same error, and it was because I was using numpy 1.13
It looks like GPyOpt does not get well with the new version of numpy.

Just downgrade your numpy version ;-)

[javiergonzalezh]

That's a great input. Thanks for commenting. Do you know exactly what is failing when trying to use 1.13?That would help to update the code in that direction as I guess that more people will have the same problem in the future. Javier

…

On 13 June 2017 at 13:56, Alberto Rubio ***@***.***> wrote: @himat <https://github.com/himat> I got the same error, and it was because I was using numpy 1.13 It looks like GPyOpt does not get well with the new version of numpy. Just downgrade your numpy version ;-) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#96 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGiS88iH6Gs-4_U6aNOUczspfIzOq06Bks5sDob5gaJpZM4NzY-n> .

[ghost]

Hi,

There are numerous places in the GPyOpt code where it tests for the presence of an array argument like this:
if X_batch != None:

With the new numpy his will cause the error described above for arrays, and I encountered the same error in acquisitions.LP (see below). The solution is to replace these tests with the following form:
if X_batch is not None:

A similar fix applies to tests of the form "== None"

I hope this helps. It should be fairly simple to find and repair all occurrences. Most of the existing tests already have the correct form.

Traceback (most recent call last):
  File "optimiser.py", line 102, in <module>
    process()
  File "optimiser.py", line 79, in process
    evaluator_type=evaluator_type)
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/methods/bayesian_optimization.py", line 244, in __init__
    self.run_optimization(max_iter=0,verbosity=self.verbosity)
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/methods/bayesian_optimization.py", line 458, in run_optimization
    super(BayesianOptimization, self).run_optimization(max_iter = max_iter, max_time = max_time,  eps = eps, verbosity=verbosity, save_models_parameters = save_models_parameters, report_file = report_file, evaluations_file= evaluations_file, models_file=models_file)
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/core/bo.py", line 108, in run_optimization
    self.suggested_sample = self._compute_next_evaluations()
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/core/bo.py", line 186, in _compute_next_evaluations
    return self.evaluator.compute_batch() 
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/core/evaluators/batch_local_penalization.py", line 45, in compute_batch
    new_sample = self.acquisition.optimize()
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/acquisitions/base.py", line 59, in optimize
    out = self.optimizer.optimize(f=self.acquisition_function, f_df=self.acquisition_function_withGradients)[0]
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/optimization/acquisition_optimizer.py", line 227, in optimize
    pref_f = f(arms)
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/acquisitions/LP.py", line 110, in acquisition_function
    return self._penalized_acquisition(x, self.model, self.X_batch, self.r_x0, self.s_x0)
  File "/usr/local/lib/python2.7/site-packages/GPyOpt/acquisitions/LP.py", line 86, in _penalized_acquisition
    if X_batch!=None:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

[jzpgit]

Hi,
@javiergonzalezh @himat
I suggest to use

def _save_model_parameter_values(self):
    if self.model_parameters_iterations is None:

in GPyOpt/core/bo.py line 209

It this way, the error may go.

Simple test:

a= np.array([1,2,3])
if a==None:
... print('Equal')
...
Traceback (most recent call last):
File "", line 1, in
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
if a is not None:
... print('Equal')
...
Equal

If you tried this method and it works, please let me know. I need to inform our IT team to update the package.

[javiergonzalezh]

Closing as this issue was fixed in the release 1.2.0.