notebook.community

Edit and run



In [ ]:

    
%load_ext autoreload
%autoreload 2
%matplotlib inline



In [ ]:

    
import matplotlib.pyplot as plt
import seaborn as sns; sns.set() # prettify matplotlib

import numpy as np
import sklearn.gaussian_process as gp
import GPy



In [ ]:

    
# local modules
import turbo as tb
import turbo.modules as tm
import turbo.plotting as tp
import turbo.gui.jupyter as tg



In [ ]:

    
# make deterministic
np.random.seed(100)

Function to optimize:



In [ ]:

    
xmin, xmax = -6, 6
ymin, ymax = -5, 5
x = np.linspace(xmin, xmax, 100)
y = np.linspace(ymin, ymax, 100)
f = lambda x,y: 1.5 * (np.sin(0.5*x)**2 * np.cos(y) + 0.1*x + 0.2*y) + \
    np.random.normal(0, 0.2, size=None if isinstance(x, float) else x.shape)
X, Y = np.meshgrid(x, y)
Z = f(X,Y)
best_z = np.min(Z)



In [ ]:

    
tp.surface_3D(X,Y,Z)



In [ ]:

    
bounds = [('x', xmin, xmax), ('y', ymin, ymax)]

op = tb.Optimiser(f, 'min', bounds, pre_phase_trials=5, settings_preset='default')

op.pre_phase_select = tm.random_selector()
op.acquisition = tm.EI(xi=0.01)

kernel = GPy.kern.Matern52(input_dim=2)
#kernel.variance.set_prior(GPy.priors.Gamma.from_EV(8, 6))
#kernel.lengthscale.set_prior(GPy.priors.Gamma.from_EV(1, 2))
op.surrogate = tm.GPySurrogate(model_params={'normalizer': True, 'kernel': kernel}, training_iterations=5)
'''
op.surrogate_factory = tm.SciKitGPSurrogate.Factory(gp_params=dict(
    kernel = gp.kernels.ConstantKernel(constant_value_bounds=(0.1, 5)) * gp.kernels.RBF(length_scale_bounds=(0.1, 5)) + gp.kernels.WhiteKernel(noise_level_bounds=(1e-05, 0.5)),
    #kernel = gp.kernels.ConstantKernel() * gp.kernels.RBF() + gp.kernels.WhiteKernel(),
), variable_iterations=lambda trial_num: [10,5,2][(trial_num-4) % 3])
'''
rec = tb.Recorder(op)



In [ ]:

    
tg.OptimiserProgressBar(op)
op.run(max_trials=30);



In [ ]:

    
#tp.plot_error(rec, true_best=best_z);
tp.plot_overview(rec)



In [ ]:

    
op.get_incumbent()



In [ ]:

    
tp.plot_surrogate_hyper_params_2D(rec, param_indices=(0,1), size_limits=(30,30), use_param_bounds=True);



In [ ]:

    
tp.interactive_plot_trial_1D(rec);



In [ ]:

    
tp.interactive_plot_trial_2D(rec, true_objective=f)

Try optimising the same function with random search



In [ ]:

    
ra = tb.Optimiser(f, 'min', bounds, pre_phase_trials=1, settings_preset='random_search')
recr = tb.Recorder(ra)
ra.run(max_trials=1000)



In [ ]:

    
tp.plot_error(recr, true_best=best_z);



In [ ]:

    
ra.get_incumbent()



In [ ]:

    
tp.interactive_plot_trial_2D(recr, true_objective=f)



In [ ]: