

In [ ]:

    
from __future__ import print_function, division, absolute_import

import GPy
import numpy as np
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline

import safeopt

mpl.rcParams['figure.figsize'] = (20.0, 10.0)
mpl.rcParams['font.size'] = 20
mpl.rcParams['lines.markersize'] = 20

Define a kernel and function

Here we define a kernel. The function is drawn at random from the GP and is corrupted my Gaussian noise



In [ ]:

    
# Measurement noise
noise_var = 0.05 ** 2

# Bounds on the inputs variable
bounds = [(-5., 5.), (-5., 5.)]

# Define Kernel
kernel = GPy.kern.RBF(input_dim=len(bounds), variance=2., lengthscale=1.0,
                      ARD=True)

# Initial safe point
x0 = np.zeros((1, len(bounds)))

# Generate function with safe initial point at x=0
def sample_safe_fun():
    while True:
        fun = safeopt.sample_gp_function(kernel, bounds, noise_var, 10)
        if fun([0,0], noise=False) > 0.5:
            break
    return fun

Interactive run of the algorithm

The slow part of running this is the plotting with matplotlib. Consider switching to the 2D level sets.



In [ ]:

    
# Define the objective function
fun = sample_safe_fun()

# The statistical model of our objective function
gp = GPy.models.GPRegression(x0, fun(x0), kernel, noise_var=noise_var)

# The optimization routine
opt = safeopt.SafeOptSwarm(gp, 0., bounds=bounds, threshold=0.2)
# parameter_set = safeopt.linearly_spaced_combinations(bounds, 100)
# opt = safeopt.SafeOpt(gp, parameter_set, 0., lipschitz=None, threshold=0.2)



In [ ]:

    
opt.plot(100, plot_3d=False)



In [ ]:

    
# Obtain next query point
x_next = opt.optimize()
# Get a measurement from the real system
y_meas = fun(x_next)
# Add this to the GP model
opt.add_new_data_point(x_next, y_meas)

opt.plot(100)



In [ ]:



In [ ]:



In [ ]:



In [ ]:



In [ ]: