Numpy实现MultilayerPerceptron

from __future__ import print_function, division
import numpy as np
import math
from sklearn import datasets

from mlfromscratch.utils import train_test_split, to_categorical, normalize, accuracy_score, Plot
from mlfromscratch.deep_learning.activation_functions import Sigmoid, Softmax
from mlfromscratch.deep_learning.loss_functions import CrossEntropy

class MultilayerPerceptron():
    """Multilayer Perceptron classifier. A fully-connected neural network with one hidden layer.
    Unrolled to display the whole forward and backward pass.

    Parameters:
    -----------
    n_hidden: int:
        The number of processing nodes (neurons) in the hidden layer. 
    n_iterations: float
        The number of training iterations the algorithm will tune the weights for.
    learning_rate: float
        The step length that will be used when updating the weights.
    """
    def __init__(self, n_hidden, n_iterations=3000, learning_rate=0.01):
        self.n_hidden = n_hidden
        self.n_iterations = n_iterations
        self.learning_rate = learning_rate
        self.hidden_activation = Sigmoid()
        self.output_activation = Softmax()
        self.loss = CrossEntropy()

    def _initialize_weights(self, X, y):
        n_samples, n_features = X.shape
        _, n_outputs = y.shape
        # Hidden layer
        limit   = 1 / math.sqrt(n_features)
        self.W  = np.random.uniform(-limit, limit, (n_features, self.n_hidden))
        self.w0 = np.zeros((1, self.n_hidden))
        # Output layer
        limit   = 1 / math.sqrt(self.n_hidden)
        self.V  = np.random.uniform(-limit, limit, (self.n_hidden, n_outputs))
        self.v0 = np.zeros((1, n_outputs))

    def fit(self, X, y):

        self._initialize_weights(X, y)

        for i in range(self.n_iterations):

            # ..............
            #  Forward Pass
            # ..............

            # HIDDEN LAYER
            hidden_input = X.dot(self.W) + self.w0
            hidden_output = self.hidden_activation(hidden_input)
            # OUTPUT LAYER
            output_layer_input = hidden_output.dot(self.V) + self.v0
            y_pred = self.output_activation(output_layer_input)

            # ...............
            #  Backward Pass
            # ...............

            # OUTPUT LAYER
            # Grad. w.r.t input of output layer
            grad_wrt_out_l_input = self.loss.gradient(y, y_pred) * self.output_activation.gradient(output_layer_input)
            grad_v = hidden_output.T.dot(grad_wrt_out_l_input)
            grad_v0 = np.sum(grad_wrt_out_l_input, axis=0, keepdims=True)
            # HIDDEN LAYER
            # Grad. w.r.t input of hidden layer
            grad_wrt_hidden_l_input = grad_wrt_out_l_input.dot(self.V.T) * self.hidden_activation.gradient(hidden_input)
            grad_w = X.T.dot(grad_wrt_hidden_l_input)
            grad_w0 = np.sum(grad_wrt_hidden_l_input, axis=0, keepdims=True)

            # Update weights (by gradient descent)
            # Move against the gradient to minimize loss
            self.V  -= self.learning_rate * grad_v
            self.v0 -= self.learning_rate * grad_v0
            self.W  -= self.learning_rate * grad_w
            self.w0 -= self.learning_rate * grad_w0

    # Use the trained model to predict labels of X
    def predict(self, X):
        # Forward pass:
        hidden_input = X.dot(self.W) + self.w0
        hidden_output = self.hidden_activation(hidden_input)
        output_layer_input = hidden_output.dot(self.V) + self.v0
        y_pred = self.output_activation(output_layer_input)
        return y_pred


def main():
    data = datasets.load_digits()
    X = normalize(data.data)
    y = data.target

    # Convert the nominal y values to binary
    y = to_categorical(y)

    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, seed=1)

    # MLP
    clf = MultilayerPerceptron(n_hidden=16,
        n_iterations=1000,
        learning_rate=0.01)

    clf.fit(X_train, y_train)
    y_pred = np.argmax(clf.predict(X_test), axis=1)
    y_test = np.argmax(y_test, axis=1)

    accuracy = accuracy_score(y_test, y_pred)
    print ("Accuracy:", accuracy)

    # Reduce dimension to two using PCA and plot the results
    Plot().plot_in_2d(X_test, y_pred, title="Multilayer Perceptron", accuracy=accuracy, legend_labels=np.unique(y))

if __name__ == "__main__":
    main()

  

 

  
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
 

文章来源: wanghao.blog.csdn.net，作者：AI浩，版权归原作者所有，如需转载，请联系作者。

原文链接：wanghao.blog.csdn.net/article/details/121558344