adding compare.py and target.py

compare.py

+import sys
+sys.path.append(r'/home/shamith/HyperSphere/')
+sys.path.append(r'/home/shamith/HyperSphere/HyperSphere/')
+sys.path.append(r'/home/shamith/HyperSphere/HyperSphere/BO/')
+
+from matplotlib import pyplot as my_plt
+from matplotlib import patches as my_patch
+import numpy as np
+import pickle
+import json
+import pandas as pd
+
+
+data_config_filename = r'/home/shamith/HyperSphere/HyperSphere/experiments/neg_birdy_D2_1/data_config.pkl'
+filename = r'/home/shamith/BayesianOptimization/logs/<function birdy at 0x7f3ca17abb70>.json'
+rename = r'/home/shamith/BayesianOptimization/plots/<function birdy at 0x7f3ca17abb70>.json'
+
+# HyperSphere
+data_config_file = open(data_config_filename, 'rb')
+for key, val in pickle.load(data_config_file).items():
+    if key == 'output':
+        y1 = val.data.numpy()
+    
+    if key == 'x_input':
+        x1 = val.data.numpy()
+
+data_config_file.close()
+
+# fmfn/BayesianOptimization
+with open(filename, "r") as f:
+    file_data = f.read()
+
+with open(filename, "r") as f:
+    lines = f.readlines()
+    first_line = lines[0]
+    file_lines = [''.join([',', line.strip(), '\n']) for line in lines[1:]]
+
+with open(rename, "w+") as f:
+    f.write('{' + '\n' + '  ' + '"logs"' + ':' + ' ' + '[' + '\n')
+    f.writelines(first_line)
+    f.writelines(file_lines)
+    f.write("    " + "]" + "\n" + "}")
+
+with open(rename) as json_file:
+    data = json.load(json_file)
+
+x2 = np.arange(1, len(lines)+1, 1)
+y2 = pd.DataFrame(data['logs']).to_numpy()[:,0]
+
+my_plt.figure('HyperSphere vs fmfn')
+my_plt.title("function = 'bird Function' :: evaluations = 50")
+red = my_patch.Patch(color='red', label='HyperSphere')
+blue = my_patch.Patch(color='blue', label='fmfn')
+my_plt.legend(handles=[red, blue])
+
+my_plt.plot(x1[:,0], y1, 'r')
+my_plt.plot(x2, y2, 'b')
+
+my_plt.show()
\ No newline at end of file

target.py

+import torch
+import math
+import numpy as np
+
+# HyperSphere Functions
+
+def neg_birdy(x):
+	flat = x.dim() == 1
+	if flat:
+		x = x.view(1, -1)
+	ndim = x.size(1)
+	n_repeat = ndim // 2
+	x = x * 2 * math.pi
+	output = 0
+	for i in range(n_repeat):
+		output += (x[:, 2 * i] - x[:, 2 * i + 1]) ** 2 + torch.exp((1 - torch.sin(x[:, 2 * i])) ** 2) * torch.cos(x[:, 2 * i + 1]) + torch.exp((1 - torch.cos(x[:, 2 * i + 1])) ** 2) * torch.sin(x[:, 2 * i])
+	output /= float(n_repeat)
+	if flat:
+		return -1.0*output.squeeze(0)
+	else:
+		return -1.0*output
+
+neg_birdy.dim = 0
+
+def bf(x):   # bohachevsky_function
+    return x[0]**2 + 2*(x[1]**2) - 0.3*torch.cos(3*math.pi*x[0]) - 0.4*torch.cos(4*math.pi*x[1]) + 0.7
+
+bf.dim = 0
+
+def nbf(x):   # bohachevsky_function
+    return -1.0*(x[0]**2 + 2*(x[1]**2) - 0.3*torch.cos(3*math.pi*x[0]) - 0.4*torch.cos(4*math.pi*x[1]) + 0.7)
+
+nbf.dim = 0
+
+def sphere(x):
+	return x[0]**2 + x[1]**2 + x[2]**2
+
+sphere.dim = 0
+
+def neg_sphere(x):
+	return -1.0*(x[0]**2 + x[1]**2 + x[2]**2)
+
+neg_sphere.dim = 0
+
+def cit(x):
+	return -0.0001*(torch.pow((torch.sin(x[0]) * torch.sin(x[1]) * torch.exp(100 - (torch.sqrt(x[0]**2 + x[1]**2)/math.pi)) + 1), 0.1))
+
+cit.dim = 0
+
+def neg_cit(x):
+	return 0.0001*(torch.pow((torch.sin(x[0]) * torch.sin(x[1]) * torch.exp(100 - (torch.sqrt(x[0]**2 + x[1]**2)/math.pi)) + 1), 0.1))
+
+neg_cit.dim = 0
+
+def neg_htf(x):
+	return torch.abs(torch.sin(x[0]) * torch.sin(x[1]) * torch.exp(torch.abs(1 - (torch.sqrt(x[0]**2 + x[1]**2)/math.pi))))
+
+neg_htf.dim = 0
+
+def htf(x):
+	return -1.0*torch.abs(torch.sin(x[0]) * torch.sin(x[1]) * torch.exp(torch.abs(1 - (torch.sqrt(x[0]**2 + x[1]**2)/math.pi))))
+
+htf.dim = 0
+
+def pf(x):
+	return -1.0*((x[0]**2)/2 + (x[1]**2/2))
+
+pf.dim = 0
+
+def neg_pf(x):
+	return ((x[0]**2)/2 + (x[1]**2/2))
+
+neg_pf.dim = 0
+
+def neg_branin(x):
+	flat = x.dim() == 1
+	if flat:
+		x = x.view(1, -1)
+	ndim = x.size(1)
+	n_repeat = ndim // 2  # changed from float to int
+	n_dummy = ndim % 2
+
+	shift = torch.cat([torch.FloatTensor([2.5, 7.5]).repeat(n_repeat), torch.zeros(n_dummy)])
+
+	if hasattr(x, 'data'):
+		x.data = x.data * 7.5 + shift.type_as(x.data)
+	else:
+		x = x * 7.5 + shift.type_as(x)
+	a = 1
+	b = 5.1 / (4 * math.pi ** 2)
+	c = 5.0 / math.pi
+	r = 6
+	s = 10
+	t = 1.0 / (8 * math.pi)
+	output = 0
+	for i in range(n_repeat):
+		output += a * (x[:, 2 * i + 1] - b * x[:, 2 * i] ** 2 + c * x[:, 2 * i] - r) ** 2 + s * (1 - t) * torch.cos(x[:, 2 * i]) + s
+	output /= float(n_repeat)
+	if flat:
+		return -1.0*output.squeeze(0)
+	else:
+		return -1.0*output
+
+neg_branin.dim = 0
+
+def mcf(x):
+	return (torch.sin(x[0] + x[1]) + (x[0] - x[1])**2 - 1.5*x[0] + 2.5*x[1] + 1)
+
+mcf.dim = 0
+
+def neg_mcf(x):
+	return -1.0*(torch.sin(x[0] + x[1]) + (x[0] - x[1])**2 - 1.5*x[0] + 2.5*x[1] + 1)
+
+neg_mcf.dim = 0
+
+def sq(x):
+	return torch.sqrt(1 - x[0]**2)
+
+sq.dim = 0
+
+# FmFn Functions
+
+def sphere(x1, x2):
+	return (x1**2 + x2**2)
+
+sphere.pbounds = {'x1': (-5.12, 5.12), 'x2': (-5.12, 5.12)}
+
+def branin(x1, x2):
+	a = 1
+	b = 5.1 / (4 * math.pi ** 2)
+	c = 5.0 / math.pi
+	r = 6
+	s = 10
+	t = 1.0 / (8 * math.pi)
+
+	return a * (x2 - b * x1**2 + c * x1 - r)**2 + s * (1 - t) * np.cos(x1) + s
+
+branin.pbounds = {'x1': (-5, 10), 'x2': (0, 15)}
+
+def neg_branin(x1, x2):
+	a = 1
+	b = 5.1 / (4 * math.pi ** 2)
+	c = 5.0 / math.pi
+	r = 6
+	s = 10
+	t = 1.0 / (8 * math.pi)
+
+	return -1.0 * (a * (x2 - b * x1**2 + c * x1 - r)**2 + s * (1 - t) * np.cos(x1) + s)
+
+neg_branin.pbounds = {'x1': (-5, 10), 'x2': (0, 15)}
+
+def birdy(x1, x2):
+	return (np.sin(x1) * np.exp((1 - np.cos(x2))**2) + np.cos(x2) * np.exp((1 - np.sin(x1)**2)) + (x1 - x2)**2) 
+
+birdy.pbounds = {'x1': (-2*np.pi, 2*np.pi), 'x2': (-2*np.pi, 2*np.pi)}
+
+def neg_birdy(x1, x2):
+	return -1.0 * (np.sin(x1) * np.exp((1 - np.cos(x2))**2) + np.cos(x2) * np.exp((1 - np.sin(x1)**2)) + (x1 - x2)**2) 
+
+neg_birdy.pbounds = {'x1': (-2*np.pi, 2*np.pi), 'x2': (-2*np.pi, 2*np.pi)}