ValueError: iterator is too large

[smdubarry]

Optimizing in 10-dimensional space, after telling about 460 or so points, I am getting this error:

>>> res = opt.tell((948.15646, -617.33538, -597.3366, -234.28106, -311.52753, -61.03256, 56.78082, -27.60636, -304.24363, 833.75898), -7.97135426448522) Traceback (most recent call last): File "<pyshell#2>", line 1, in <module> res = opt.tell((948.15646, -617.33538, -597.3366, -234.28106, -311.52753, -61.03256, 56.78082, -27.60636, -304.24363, 833.75898), -7.97135426448522) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\skopt\optimizer\optimizer.py", line 443, in tell return self._tell(x, y, fit=fit) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\skopt\optimizer\optimizer.py", line 502, in _tell acq_func_kwargs=self.acq_func_kwargs) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\skopt\acquisition.py", line 50, in _gaussian_acquisition func_and_grad = gaussian_ei(X, model, y_opt, xi, return_grad) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\skopt\acquisition.py", line 276, in gaussian_ei mu, std = model.predict(X, return_std=True) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\skopt\learning\gaussian_process\gpr.py", line 324, in predict y_var -= np.einsum("ki,kj,ij->k", K_trans, K_trans, K_inv) File "C:\Users\smdubarry\AppData\Local\Programs\Python\Python37-32\lib\site-packages\numpy\core\einsumfunc.py", line 1346, in einsum return c_einsum(*operands, **kwargs) ValueError: iterator is too large

Initial optimizer is:

opt = Optimizer(dimensions = [(-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0), (-1000.0, 1000.0)], acq_func = "EI", n_random_starts = 9)

Am I just trying to do too much?

[chnzhangrui]

Hi,

Yes, it is definitely because it couldn't handle the Gaussian process. Computing Gaussian process gets more complicated when the known points increases - it updates the covariance matrix based on all past measurement, so will get slower and slower when more measurements are available.
Personally I would suggest you divide your ten-dimensional space to orthogonal subspaces. If you know some hyperparameters don't reply on others, you could optimise them separately. Then use fewer iterations.