PythonCodes/perceptron.py at master · RScrusoe/PythonCodes · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
import numpy as np
from matplotlib import pyplot as plt
import csv

# np.random.seed(10)
slope = 0.3
intercept = -0.1
epochs = 5

def data_load(file):
    f = open(file)
    r = csv.reader(f)
    d = {}
    for i in r:
        i[0] = int(i[0])
        if i[0] not in d:
            d[i[0]] = []
            d[i[0]].append([float(i[1]),float(i[2])])
        else:
            d[i[0]].append([float(i[1]),float(i[2])])
    f.close()
    return d
# d = data_load('datapoints.csv')

def line_function(x):
    return slope*x + intercept


def random_data_points(n,xl,xh,yl,yh):
    l = []
    for i in range(n):
        x = np.random.uniform(xl,xh)
        y = np.random.uniform(yl,yh)
        lineY = line_function(x)
        if y >= lineY:
            label = +1
        else:
            label = -1
        l.append([label,x,y])
    return l

xy_data = random_data_points(100,-1,1,-1,1)


class Perceptron():
    n = 3
    weights = []
    learning_rate = 0.1
    bias = 1

    def __init__(self,n):
        n = n
        for i in range(self.n):
            self.weights.append(np.random.uniform(-1,1))

    def guess(self,inputs):
        all_sum = 0
        for i in range(self.n):
            all_sum += inputs[i]*self.weights[i]
        return np.sign(all_sum)

    def train(self,inp,target):
        guess = self.guess(inp)
        error = target - guess
        for i in range(self.n):
            self.weights[i] += (error * inp[i] * self.learning_rate)


    def draw_trained_line(self):
        x1 = -1
        x2 = 1
        y1 = -1*(self.weights[2]/self.weights[1]) -1*(self.weights[0]/self.weights[1])*(x1)
        y2 = -1*(self.weights[2]/self.weights[1]) -1*(self.weights[0]/self.weights[1])*(x2)
        print(-1*(self.weights[0]/self.weights[1]),-1*(self.weights[2]/self.weights[1]))
        plt.plot([x1,x2],[y1,y2],ls='dashed',c='r')


p = Perceptron(3)

#setting axes
plt.axis([-1,1,-1,1])

# Plotting Acrual line
plt.plot([-1,1],[line_function(-1),line_function(1)],ls='solid')


# Training for Epochs
# For each epoch: Train, check score
# For last epoch plot accordingly
for epoch in range(epochs):
    result = 0
    for i in xy_data:
        inp = i[1:] + [1]
        out = p.train(inp,i[0])
    for i in xy_data:
        inp = i[1:] + [1]
        guess = p.guess(inp)
        if guess == i[0]:
            result += 1

    if epoch == (epochs - 1):
        result = 0
        for i in xy_data:
            inp = i[1:] + [1]
            guess = p.guess(inp)
            if i[0] == 1:
                m = '*'
            else:
                m = 'o'
            if guess == i[0]:
                result += 1
                c = 'g'
            else:
                c = 'r'
            plt.scatter(i[1],i[2],marker=m,c = c)
            # plt.pause(0.0001)
    print(result)


p.draw_trained_line()
plt.show()


# print(p.weights)