Question

我意识到这并不是我要实现的神经网络，而是更多的一层感知器。我以为这很容易（我想是这样，但我只是想不通。我编写的代码如下：

import numpy as np
import pylab
import matplotlib.pyplot as plt

from sklearn.utils import shuffle

def shuffling(X,t):
    if X.shape[0] != t.shape[0]:
        X = np.transpose(X)
        X,t = shuffle(X,t)
        X = np.transpose(X)
        t = np.transpose(t)
    else:
        X,t = shuffle(X,t)
        X = np.transpose(X)
        t = np.transpose(t)
    return X,t

def generating_data():
    n = 100
    mA = [-1.0, 2.5]
    sigmaA = 0.5
    mB = [1.5, 0.9]
    sigmaB = 0.5
    A_1 = np.random.normal(mA[0], sigmaA, n)
    A_2 = np.random.normal(mA[1], sigmaA, n)
    classA = np.stack((A_1, A_2), axis=1)
    B_1 = np.random.normal(mB[0], sigmaB, n)
    B_2 = np.random.normal(mB[1], sigmaB, n)
    classB = np.stack((B_1,B_2), axis=1)
    tot_class = np.vstack((classA, classB))
    target_values_A = np.ones(n)
    target_values_B = np.ones(n)*(-1)
    tot_target = np.concatenate((target_values_A, target_values_B), axis=0)
    X = np.insert(tot_class, 2, values = 1, axis =1)
    X, t = shuffling(X, tot_target)
    return X,t, classA, classB, tot_class


X,t, classA, classB, tot_class = generating_data()

def perceptron2(X, t, max_iter):
    n = 200
    lr = 0.5
    w = np.random.normal(0, 1/n, (1,3))
    for k in range(max_iter):
        y = np.dot(w, X)
        t = np.reshape(t,(1,200))
        w = w - lr*np.dot((t-y),np.transpose(X))
        #X,t = shuffling(X,t)
    return w


w = perceptron2(X,t,10)
print(w)


#print(w)

def plotting(classA, classB, tot_class):
    w = perceptron2(X,t,100)
    area = np.pi*3
    plt.scatter(classA[:,0], classA[:,1], s = area, c='b', alpha=0.5)
    plt.scatter(classB[:,0], classB[:,1],s=area, c= 'r', alpha = 0.5)
    grid = np.linspace(np.amin(tot_class),np.amax(tot_class))
    y = []
    for x in grid:[enter image description here][1]
        slope = -(w[0][0] / w[0][2]) / (w[0][0] / w[0][1])
        intercept = -w[0][0] / w[0][2]
        print(slope)
        print(intercept)
        y.append((slope*x) + intercept)
    plt.plot(grid,y, color = 'black')

    plt.show()

plotting(classA,classB,tot_class)

这是一个二进制分类问题，正如您从代码中看到的那样，我仅使用了两个功能。对于缺少图像感到抱歉，但是如果您自行运行，您将看到超平面完全关闭。从这一点上可以使我前进的任何事情都受到赞赏，谢谢

感知器，使用增量规则

0 个答案: