自适应线性神经网络Adaline的python实现详解

自适应线性神经网络Adaptive linear network， 是神经网络的入门级别网络。

相对于感知器，采用了f（z）=z的激活函数，属于连续函数。

代价函数为LMS函数，最小均方算法，Least mean square。

实现上，采用随机梯度下降，由于更新的随机性，运行多次结果是不同的。

'''
Adaline classifier

created on 2019.9.14
author: vince
'''
import pandas
import math
import numpy
import logging
import random
import matplotlib.pyplot as plt

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

'''
Adaline classifier

Attributes
w: ld-array = weights after training
l: list = number of misclassification during each iteration
'''
class Adaline:
 def __init__(self, eta = 0.001, iter_num = 500, batch_size = 1):
   '''
   eta: float = learning rate (between 0.0 and 1.0).
   iter_num: int = iteration over the training dataset.
   batch_size: int = gradient descent batch number,
     if batch_size == 1, used SGD;
     if batch_size == 0, use BGD;
     else MBGD;
   '''

self.eta = eta;
   self.iter_num = iter_num;
   self.batch_size = batch_size;

def train(self, X, Y):
   '''
   train training data.
   X:{array-like}, shape=[n_samples, n_features] = Training vectors,
     where n_samples is the number of training samples and
     n_features is the number of features.
   Y:{array-like}, share=[n_samples] = traget values.
   '''
   self.w = numpy.zeros(1 + X.shape[1]);
   self.l = numpy.zeros(self.iter_num);
   for iter_index in range(self.iter_num):
     for rand_time in range(X.shape[0]):
       sample_index = random.randint(0, X.shape[0] - 1);
       if (self.activation(X[sample_index]) == Y[sample_index]):
         continue;
       output = self.net_input(X[sample_index]);
       errors = Y[sample_index] - output;
       self.w[0] += self.eta * errors;
       self.w[1:] += self.eta * numpy.dot(errors, X[sample_index]);
       break;
     for sample_index in range(X.shape[0]):
       self.l[iter_index] += (Y[sample_index] - self.net_input(X[sample_index])) ** 2 * 0.5;
     logging.info("iter ％s: w0(％s), w1(％s), w2(％s), l(％s)" ％
         (iter_index, self.w[0], self.w[1], self.w[2], self.l[iter_index]));
     if iter_index > 1 and math.fabs(self.l[iter_index - 1] - self.l[iter_index]) < 0.0001:
       break;

def activation(self, x):
   return numpy.where(self.net_input(x) >= 0.0 , 1 , -1);

def net_input(self, x):
   return numpy.dot(x, self.w[1:]) + self.w[0];

def predict(self, x):
   return self.activation(x);

def main():
 logging.basicConfig(level = logging.INFO,
     format = '％(asctime)s ％(filename)s[line:％(lineno)d] ％(levelname)s ％(message)s',
     datefmt = '％a, ％d ％b ％Y ％H:％M:％S');

iris = load_iris();

features = iris.data[:99, [0, 2]];
 # normalization
 features_std = numpy.copy(features);
 for i in range(features.shape[1]):
   features_std[:, i] = (features_std[:, i] - features[:, i].mean()) / features[:, i].std();

labels = numpy.where(iris.target[:99] == 0, -1, 1);

# 2/3 data from training, 1/3 data for testing
 train_features, test_features, train_labels, test_labels = train_test_split(
     features_std, labels, test_size = 0.33, random_state = 23323);

logging.info("train set shape:％s" ％ (str(train_features.shape)));

classifier = Adaline();

classifier.train(train_features, train_labels);

test_predict = numpy.array([]);
 for feature in test_features:
   predict_label = classifier.predict(feature);
   test_predict = numpy.append(test_predict, predict_label);

score = accuracy_score(test_labels, test_predict);
 logging.info("The accruacy score is: ％s "％ (str(score)));

#plot
 x_min, x_max = train_features[:, 0].min() - 1, train_features[:, 0].max() + 1;
 y_min, y_max = train_features[:, 1].min() - 1, train_features[:, 1].max() + 1;
 plt.xlim(x_min, x_max);
 plt.ylim(y_min, y_max);
 plt.xlabel("width");
 plt.ylabel("heigt");

plt.scatter(train_features[:, 0], train_features[:, 1], c = train_labels, marker = 'o', s = 10);

k = - classifier.w[1] / classifier.w[2];
 d = - classifier.w[0] / classifier.w[2];

plt.plot([x_min, x_max], [k * x_min + d, k * x_max + d], "go-");

plt.show();

if __name__ == "__main__":
 main();

来源：https://www.cnblogs.com/thsss/p/11520673.html