Handwriting a machine learning entry algorithm-Perceptron algorithm

100 points were made using 4x+5y=2000 as the dividing line;

The initial dividing line is 0, 0;

After 1000 rounds of correction, the result is:

X+31 y = 11876

Comparison results 4 x + 5 y = 2000

is still relatively close.

Just beginning to update w the line of code mistaken, thought is to use predict to correct, in fact, should use the real value of sample to correct.

Import random;

def find_split (points):
    w= (0,0,0)
    For _ in range (1,2000):
        print ' w= ' +str (w);
        For PT in points:
            (x1,x2,z) = pt;
            (W1,W2,W3) =w;
            predict = Int ((W1+W2*X1+W3*X2) >0) *2-1
            If Predict!=z:
                print ' Wrong: ' +str (PT)
                W= (W1+Z,W2+Z*X1,W3+Z*X2);
                # break;
            Else
                print ' Right: ' +str (PT)
    Return w;

def test_split (points,w):
    Points_2 = Filter (Lambda pt: ((int (w[0]+w[1]*pt[0]+w[2]*pt[1) >=0) *2-1) ==pt[2],points)
    return points_2;

def init_points (max_x,max_y,num_of_pts):
    Points=[];
    For I in Range (1,num_of_pts,1):
        x = Int (random.random () *max_x);
        y = Int (random.random () *max_y);
        z = Int ((4*x+5*y) >=2000) *2-1
        Points.append ((x, Y, z));
    return points;

if __name__ = = ' __main__ ':
    Points = init_points (400,500,100);
    Print points;
    Line = Find_split (points);
    print (line);
    pts = Test_split (points,line);
    Print points;
    Print Len (pts);

Handwriting a machine learning entry algorithm-Perceptron algorithm