Classification in Action:Enter the Perceptron - a primitive kind of SVM algorithm

Author: DevOpsThinh

ml/supervised_learning/perceptrons/perceptron_simplest_way_to_find_a_hyperplane_for_classification.py

@@ -0,0 +1,71 @@
+# Learner: Nguyen Truong Thinh
+# Contact me: [email protected] || +84393280504
+#
+# Topic: Supervised Learning: Classifications in Action.
+#           A primitive kind of support vector machine - Perceptron
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+
+# Can try two different arrangements of x
+# x = np.array([[1,2], [3,6], [4,7], [5,6], [1,3], [2.5,4], [2,3]])
+x = np.array([[5, 2], [3, 6], [2, 7], [1, 6], [5, 3], [3.5, 4], [4, 3]])
+y_orig = np.array([-1, 1, 1, 1, -1, -1, -1])
+plt.scatter(x[:, 0], x[:, 1], c=y_orig)
+plt.show()
+
+# Length of y(input) vector
+L = y_orig.size
+# Separate X1, X2 to see algorithm clearly
+X1 = x[:, 0].reshape(L, 1)
+X2 = x[:, 1].reshape(L, 1)
+y = y_orig.reshape(L, 1)
+# Creating our weights, start at 0 for 1st iteration
+w0, w1, w2 = 1, 0, 0
+# Counter to go through each point
+count = 0
+iteration = 1
+# Learning rate
+alpha = 0.01
+
+while iteration < 1000:
+    y_hat = w0 + X1 * w1 + X2 * w2
+    prod = y_hat * y  # If > 1, the prediction is correct
+    for p in prod:
+        if p <= 1:
+            # Nudge w vector in right direction
+            w0 = w0 + alpha * y[count]
+            w1 = w1 + alpha * y[count] * X1[count]
+            w2 = w2 + alpha * y[count] * X2[count]
+
+        count += 1
+    count = 0
+    iteration += 1
+
+print('w0', w0)
+print('w1', w1)
+print('w2', w2)
+
+# Final perceptron answers (If < 0 -> category 1, else > 0 -> category 2)
+y = w0 + X1 * w1 + X2 * w2
+# Plot the predictions via Perceptron algorithm
+plt.scatter(x[:, 0], x[:, 1], c=(y < 0).reshape(1, -1)[0])
+# Plot the line
+q = np.array([0, 7])  # 2 points on x axis.
+# Calculated hyperplane (a line)
+x_ = -(w1 / w2).reshape(1, -1) * q - w0 / w2
+# f(x) = w.x+b+1 support vector line
+x_p = -(w1 / w2).reshape(1, -1) * q - w0 / w2 - 1 / w2
+# f(x) = w.x+b+1 support vector line
+x_n = -(w1 / w2).reshape(1, -1) * q - w0 / w2 + 1 / w2
+
+plt.plot(q, x_[0])
+plt.plot(q, x_p[0], 'r--')
+plt.plot(q, x_n[0], 'r--')
+plt.xlim([0, 6])
+plt.ylim([0, 12])
+plt.grid()
+plt.xlabel(r'$X_{1}$')
+plt.ylabel(r'$X_{2}$')
+plt.show()