def generate_data():
size_clusterA = no_of_points // clusterA
size_clusterB = no_of_points // clusterB
colorA = ['#2AFB05', '#4CD933', '#56CF41', '#56CF41', '#27C40B', '#71FB58']
colorB = ['#DD3E28', '#BB6B20', '#E518B4', '#B4703B', '#FB10A2', '#AA5B43']
for i in range(clusterA):
tmp = []
if i != clusterA - 1:
tmp = randomFunc.generate_N_random(
[x + i * multiplication_ratio for x in mean_classA],
sigma_classA, 2, size_clusterA)
else:
tmp = randomFunc.generate_N_random(
[x + i * multiplication_ratio for x in mean_classA],
sigma_classA, 2, no_of_points - size_clusterA * i)
naive.add_plotting_data(tmp, 'ro', colorA[i],
"Class A Shade " + str(i + 1))
for i in range(clusterB):
tmp = []
if i != clusterB - 1:
tmp = randomFunc.generate_N_random(
[x + i * multiplication_ratio for x in mean_classB],
sigma_classB, 2, size_clusterB)
else:
tmp = randomFunc.generate_N_random(
[x + i * multiplication_ratio for x in mean_classB],
sigma_classB, 2, no_of_points - size_clusterB * i)
naive.add_plotting_data(tmp, 'ro', colorB[i],
"Class B Shade " + str(i + 1))
naive.plot_curve("GMM", [-1, 6, -1, 6], ["X Axis", "Y Axis"])
def generate_GMM_decision_boundary():
pos = []
neg = []
for i in np.arange(-1, 6, 0.01):
for j in np.arange(-1, 6, 0.01):
diff = get_diff_of_dscriminant_function(i, j)
if diff >= 0:
pos.append([i, j])
else:
neg.append([i, j])
naive.add_plotting_data(pos, '.', 'black', "Region A")
naive.add_plotting_data(neg, '.', 'white', "Region B")
generate_data()
def single_update_perceptron_learning():
global c1_x, c1_y, c1_z, c2_x, c2_y, c2_z
global error_log_single
error_log_single = []
itr = 0
w = np.zeros((4, 1))
x = np.zeros((4, 1))
plt.ion()
total_iterations = 0
while itr < no_of_data_points:
x[0][0] = c1_x[itr]
x[1][0] = c1_y[itr]
x[2][0] = c1_z[itr]
x[3][0] = 1
if (np.matmul(np.transpose(w), x)) <= 0:
w = w + neta * x
itr = -1
x[0][0] = c2_x[0 if itr == -1 else itr]
x[1][0] = c2_y[0 if itr == -1 else itr]
x[2][0] = c2_z[0 if itr == -1 else itr]
x[3][0] = 1
if (np.matmul(np.transpose(w), x)) > 0:
w = w - neta * x
itr = -1
check_error(w, total_iterations)
itr += 1
total_iterations += 2
x1 = []
y1 = []
z1 = []
for i in np.arange(-2, 7, 0.1):
for j in np.arange(-2, 7, 0.1):
x1.append(i)
y1.append(j)
z = (w[0][0] * i + w[1][0] * j + w[3][0]) / (-1 * w[2][0])
z1.append(z)
# Plotting Curves
data_classA = []
data_classB = []
data = []
for i in range(len(c1_x)):
data_classA.append([c1_x[i], c1_y[i], c1_z[i]])
for i in range(len(c2_x)):
data_classB.append([c2_x[i], c2_y[i], c2_z[i]])
for i in range(len(x1)):
data.append([x1[i], y1[i], z1[i]])
add_plotting_data(data_classA, 'o', 'red', "Class A")
add_plotting_data(data_classB, 'o', 'blue', "Class B")
add_plotting_data(data, 'None', 'green', "Plane")
plot_3d_curve("Single Update Perceptron Learning")
def batch_update_perceptron_learning():
global c1_x, c1_y, c1_z, c2_x, c2_y, c2_z
global error_log_batch
error_log_batch = []
itr = 0
w = np.zeros((4, 1))
x = np.zeros((4, 1))
plt.ion()
check = False
total_iterations = 0
while check != True:
itr = 0
error_points = []
check = True
while itr < no_of_data_points:
x[0][0] = c1_x[itr]
x[1][0] = c1_y[itr]
x[2][0] = c1_z[itr]
x[3][0] = 1
if (np.matmul(np.transpose(w), x)) <= 0:
error_points.append([c1_x[itr], c1_y[itr], c1_z[itr], True])
check = False
x[0][0] = c2_x[itr]
x[1][0] = c2_y[itr]
x[2][0] = c2_z[itr]
x[3][0] = 1
if (np.matmul(np.transpose(w), x)) > 0:
error_points.append([c2_x[itr], c2_y[itr], c2_z[itr], False])
check = False
itr += 1
error_log_batch.append([total_iterations, len(error_points)])
for point in error_points:
x[0][0] = point[0]
x[1][0] = point[1]
x[2][0] = point[2]
x[3][0] = 1
if point[3]:
w = w + neta * x
else:
w = w - neta * x
total_iterations += 1
# Plotting Curves
x1 = []
y1 = []
z1 = []
data_classA = []
data_classB = []
data = []
for i in np.arange(-2, 7, 0.1):
for j in np.arange(-2, 7, 0.1):
x1.append(i)
y1.append(j)
z = (w[0][0] * i + w[1][0] * j + w[3][0]) / (-1 * w[2][0])
z1.append(z)
for i in range(len(c1_x)):
data_classA.append([c1_x[i], c1_y[i], c1_z[i]])
for i in range(len(c2_x)):
data_classB.append([c2_x[i], c2_y[i], c2_z[i]])
for i in range(len(x1)):
data.append([x1[i], y1[i], z1[i]])
add_plotting_data(data_classA, 'o', 'red', "Class A")
add_plotting_data(data_classB, 'o', 'blue', "Class B")
add_plotting_data(data, 'None', 'green', "Plane")
plot_3d_curve("Batch Update Perceptron Learning")
data_classA = []
data_classB = []
data = []
for i in np.arange(-2, 7, 0.1):
for j in np.arange(-2, 7, 0.1):
x1.append(i)
y1.append(j)
z = (w[0][0] * i + w[1][0] * j + w[3][0]) / (-1 * w[2][0])
z1.append(z)
for i in range(len(c1_x)):
data_classA.append([c1_x[i], c1_y[i], c1_z[i]])
for i in range(len(c2_x)):
data_classB.append([c2_x[i], c2_y[i], c2_z[i]])
for i in range(len(x1)):
data.append([x1[i], y1[i], z1[i]])
add_plotting_data(data_classA, 'o', 'red', "Class A")
add_plotting_data(data_classB, 'o', 'blue', "Class B")
add_plotting_data(data, 'None', 'green', "Plane")
plot_3d_curve("Batch Update Perceptron Learning")
if __name__ == "__main__":
generate_data()
single_update_perceptron_learning()
batch_update_perceptron_learning()
add_plotting_data(error_log_single, 'None', 'blue', "Single Update")
add_plotting_data(error_log_batch, 'None', 'red', "Batch Update")
plot_curve("Error Rate", False, ["Iterations", "Misclassified Points"])
plt.show()
plt.pause(20)