Create 04_PerceotronAlgorithm.py

1.Perceotron Algorithm

Author: KingBing

04_PerceotronAlgorithm.py

@@ -0,0 +1,86 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+#dataset
+X1 = np.array([-0.62231486, -0.96251306,  0.42269922, -1.452746  , -0.66915783,
+               -0.35716016,  0.49505163, -1.8117848 ,  0.53376487, -1.86923838,
+                0.71434306, -0.4055084 ,  0.82887254,  0.81221287,  1.44280951,
+               -0.45599278, -1.16715888,  1.08913131, -1.61470741,  1.61113001,
+               -1.4532688 ,  1.04872588, -1.52312195, -1.62831727, -0.25191539])
+
+X2 = np.array([-1.67427011, -1.81046748,  1.20384694, -0.41572751,  0.66851908,
+               -1.75435288, -1.57532207, -1.22329618, -0.84375819,  0.52873296,
+               -1.10837773,  0.04612922,  0.67696196,  0.84618152, -0.77362548,
+                0.99153072,  1.7896494 , -0.38343121, -0.21337742,  0.64754817,
+                0.36719101,  0.23132427,  1.07029963,  1.62919909, -1.53920827])
+
+Y = np.array([  1.,  -1.,  -1.,  -1.,  -1.,
+                1.,   1.,  -1.,   1.,  -1.,
+                1.,  -1.,   1.,   1.,   1., 
+               -1.,  -1.,   1.,  -1.,   1., 
+               -1.,   1.,  -1.,  -1.,   1.])
+               
+xl1 = yl1 = -2.0
+xl2 = yl2 = 2.0               
+
+#Plotting distribution of a dataset
+def plot_data(filename = 'data0'):
+    fig = plt.figure()
+    ax = fig.add_subplot(111)
+    plt.scatter(X1[Y >= 0], X2[Y >= 0], s = 80, c = 'b', marker = "o")
+    plt.scatter(X1[Y <  0], X2[Y  < 0], s = 80, c = 'r', marker = "^")
+    #ax.set_xlim([-2.0, 2.0])
+    #ax.set_ylim([-2.0, 2.0])    
+    ax.set_xlim(xl1, xl2) 
+    ax.set_ylim(yl1, yl2)       
+    fig.set_size_inches(6, 6)
+    plt.show()  
+    
+#Perceotron algorithm    
+def plot_data_and_line(w1,w2):
+    w1,w2 = float(w1),float(w2)
+    if w2 != 0 :
+        y1,y2 = (-w1*(xl1))/w2, (-w1*(xl2))/w2
+        vx1,vy1 = [xl1,xl2,xl2,xl1,xl1], [y1,y2,yl2,yl2,y1]
+        vx2,vy2 = [xl1,xl2,xl2,xl1,xl1], [y1,y2,yl1,yl1,y1]
+    elif w1 != 0:
+        vx1,vy1 = [xl2,0,0,xl2,xl2], [yl1,yl1,yl2,yl2,yl1]
+        vx2,vy2 = [xl1,0,0,xl1,xl1], [yl1,yl1,yl2,yl2,yl1]
+    else:
+        print "ERROR, Invalid w1 and w2."
+        return;
+    if  w2 > 0 or ( w2 == 0 and w1 > 0):
+        c1,c2 = 'b','r'
+    else:
+        c1,c2 = 'r','b'
+    fig = plt.figure()
+    ax = fig.add_subplot(111)
+    plt.scatter(X1[Y > 0], X2[Y > 0], s = 80, c = 'b', marker = "o")
+    plt.scatter(X1[Y<= 0], X2[Y<= 0], s = 80, c = 'r', marker = "^")
+    plt.fill(vx1, vy1, c1, alpha = 0.25)
+    plt.fill(vx2, vy2, c2, alpha = 0.25)
+    ax.set_title(("w1 = %s, w2 = %s")%( w1, w2))
+    ax.set_xlim(xl1, xl2)
+    ax.set_ylim(yl1, yl2)
+    fig.set_size_inches(6, 6)
+    plt.show()    
+        
+def learn_perceptron(times=1000):
+    w1,w2 = 1,1
+    for i in range(times):
+        ERR = (w1*X1+w2*X2) * Y < 0
+        if len(filter(bool,ERR)) > 0:
+           err_x1,err_x2,err_y = X1[ERR][0],X2[ERR][0],Y[ERR][0]
+           w1,w2 = (w1+err_y*err_x1),(w2+err_y*err_x2)
+        else: 
+           print "Complete!"
+           break;
+    plot_data_and_line(w1,w2)    
+    
+    
+if __name__ == '__main__':
+  #plot_data();      
+  #plot_data_and_line(1,1);
+  learn_perceptron();
+    
+