美文网首页
深度学习准备

深度学习准备

作者: 十块腹肌的小胡子 | 来源:发表于2018-12-16 19:29 被阅读0次

import sys, os

import numpyas np

sys.path.append(os.pardir)

from dataset.mnistimport load_mnist

import matplotlib.pyplotas plt

from PILimport Image

import pickle

#图片预览

# def img_show(img):

#    pil_img = Image.fromarray(np.uint8(img)) #将numpy数组的图像数据转换为PIL的数据对象

#    pil_img.show()

# img = x_train[0]

# label = t_train[0]

# print(label)

# print(img.shape)

# img = img.reshape(28, 28)

# print(img.shape)

# img_show(img)

# def sigmoid(z):

#    return 1 / (1 + np.exp(-z))

def softmax(a):

c = np.max(a)

exp_a = np.exp(a-c)

exp_sum = np.sum(exp_a)

return exp_a / exp_sum

#数据准备

def get_data():

(x_train, t_train), (x_test, t_test) = load_mnist(normalize=True,  flatten=True, one_hot_label=False)

return x_train, t_train

#

# def init_network():

#    with open('sample_weight.pkl', 'rb') as f : #读取二进制文件,注意文件是否过大决定读取方式,此处读取不到该文件

#        network = pickle.load(f)

#    return network

# def predict(network, x):

#    W1, W2, W3 = network['W1'], network['W2'], network['W3']

#    b1, b2, b3 = network['b1'], network['b1'], network['b1']

#    z1 = np.dot(x, W1) + b1

#    a1 = sigmoid(z1)

#    z2 = np.dot(a1, W2) + b2

#    a2 = sigmoid(z2)

#    z3 = np.dot(a2, W3) + b3

#    y = softmax(z3)

#    return y

# x, t = get_data()

# network = init_network()

# accuracy_cnt = 0

# for i in range(len(x)):

#    y = predict(network, x)

#    p = np.argmax(y)

#    if p == t[i]:

#        accuracy_cnt += 1

# print('accuracy:'+ str(float(accuracy_cnt) / len(x)))

#批处理,加快运算速度

# x, t = get_data()

# network = init_network()

# batch_size = 100

# accuracy_cnt = 0

# for i in range(0, len(x), batch_size):

#    x_batch = x[i:i+batch_size]

#    y_batch = predict(network, x_batch)

#    p = np.argmax(y_batch, axis=1) #矩阵的第0维是列方向,第1维是行方向

#    accuracy_cnt += np.sum(p == t[i:i+batch_size])

# print('accuracy:'+ str(float(accuracy_cnt) / len(x)))

#第4章

#均方误差

# def mean_squared_error(y, t):

#    return 0.5 * np.sum((y-t)**2)

# t = np.array([0, 0, 1, 0, 0, 0, 0, 0, 0, 0])

# y = np.array([0.1, 0.05, 0.6, 0.0, 0.05, 0.1, 0.0, 0.1, 0.0, 0.0])

# print(mean_squared_error(y, t))

#交叉熵误差

# def cross_entropy_error(y, t):

#    delta = 1e-7

#    return -np.sum(t * np.log(y + delta))

# t = np.array([0, 0, 1, 0, 0, 0, 0, 0, 0, 0])

# y = np.array([0.1, 0.05, 0.6, 0.0, 0.05, 0.1, 0.0, 0.1, 0.0, 0.0])

# print(cross_entropy_error(y, t))

#随机抽取

# x_train, t_train = get_data()

# train_size = x_train.shape[0]

# batch_size = 10

# batch_mask = np.random.choice(train_size, batch_size)

# x_batch = x_train[batch_mask]

# t_batch = t_train[batch_mask]

# print(batch_mask)

# print(x_train)

#交叉熵误差,支持单个和批量数据

# def cross_entropy_error(y, t):

#    if y.ndim == 1:

#        t = t.reshape(1,t. size)

#        y = y.reshape(1, y.size)

#    batch_size = y.shape[0]

#    return -np.sum(t * np.log(y + 1e-7)) / batch_size

# #如果监督数据是标签形式,计算交叉熵误差

def cross_entrory_error(y, t):

if y.ndim ==1:

t = t.reshape(1, t.size)

y = y.reshape(1, y.size)

batch_size = y.shape[0]

return -np.sum(np.log(y[np.arange(batch_size), t] +1e-7)) / batch_size

#数值微分

def function_1(x):

return 0.01*x**2 +0.1*x

# # x = np.arange(0, 20, 0.1)

# # y = function_1(x)

# # plt.xlabel('x')

# # plt.ylabel('y')

# # plt.plot(x, y)

# # plt.show()

# def numerical_diff(f, x):

#    h = 1e-4

#    return (f(x+h) - f(x-h)) / (2*h)

# print(numerical_diff(function_1, 5))

#计算给定x的偏导数,例如(x1,x2,x3),会计算出三个偏导数

def numerical_gradient_nobatch(f, x):

h =1e-4

    grad = np.zeros_like(x)

for idxin range(x.size):

val = x[idx]

#f1(x)

        x[idx] = val + h

fxh1 = f(x)

#f2(x)

        x[idx] = val - h

fxh2 = f(x)

grad[idx] = (fxh1 - fxh2) / (2*h)

x[idx] = val

return grad

#包装上述偏导计算函数,如果维度大于1维,通过enumerate每行取x for循环计算

def numerical_gradient(f, X):

if X.ndim ==1:

return numerical_gradient_nobatch(f, X)

else:

grad = np.zeros_like(X)

for idx, xin enumerate(X):

grad[idx] = numerical_gradient_nobatch(f, x)

return grad

def function_2(x):

return x[0]**2 + x[1]**2

#g = numerical_gradient(function_2, np.array([3.0, 4.0])) #没有加小数点,导致结果差距很大

#使用梯度下降法求最小值

def gradient_decent(f, init_x, lr =0.01, step_num =100):

x = init_x

for iin range(step_num):

grad = numerical_gradient(f, x)

x -=lr * grad

return x

# init_x = np.array([-3.0, 4.0])

# g = gradient_decent(function_2, init_x, lr=0.1, step_num=100)

# print(g)

#神经网络的类

class simpleNet:

def __init__(self):

self.W = np.random.randn(2,3)

def predict(self, x):

return np.dot(x, self.W)

def loss(self, x, t):

z =self.predict(x)

y = softmax(z)

loss = cross_entrory_error(y, t)

return loss

net = simpleNet()

print(net.W)

x = np.array([0.6, 0.9])

p = net.predict(x)

print(p)

print(np.argmax(p))

t = np.array([0, 0, 1])

print(net.loss(x, t))

#偏导计算函数的输入变量,函数+自变量权重

f =lambda w : net.loss(x, t)

dW = numerical_gradient(f, net.W)

print(dW)

相关文章

网友评论

      本文标题:深度学习准备

      本文链接:https://www.haomeiwen.com/subject/mkbxkqtx.html