很早之前就学完深度学习第二周的课程,由于保研的事情一直心烦,做什么事情都索然无味,还有七天出结果,今天终于沉下心把剩余的部分跟着blog具有神经网络思维的Logistic回归码了一遍,老师课程讲的很清楚,自己也做了笔记,但是编程作业中还是有些语句不理解,下面贴出自己的源码和运行结果:
lr_utils.py文件
import numpy as np
import h5py
def load_dataset():
train_dataset = h5py.File('datasets/train_catvnoncat.h5', "r")
# 保存训练集里的图像数据(本训练集有209张64×64的图像)
train_set_x_orig = np.array(train_dataset["train_set_x"][:]) # your train set features
# 保存训练集的图像对应的分类值(【0|1】0表示不是猫,1表示是猫)
train_set_y_orig = np.array(train_dataset["train_set_y"][:]) # your train set labels
test_dataset = h5py.File('datasets/test_catvnoncat.h5', "r")
# 保存测试集里的图像数据(本训练集有50张64×64的图像)
test_set_x_orig = np.array(test_dataset["test_set_x"][:]) # your test set features
# 保存测试集的图像对应的分类值(【0|1】0表示不是猫,1表示是猫 )
test_set_y_orig = np.array(test_dataset["test_set_y"][:]) # your test set labels
# 保存的是以bytes类型保存的两个字符串数据,数据位:
classes = np.array(test_dataset["list_classes"][:]) # the list of classes
train_set_y_orig = train_set_y_orig.reshape((1, train_set_y_orig.shape[0]))
test_set_y_orig = test_set_y_orig.reshape((1, test_set_y_orig.shape[0]))
return train_set_x_orig, train_set_y_orig, test_set_x_orig, test_set_y_orig, classes
main.py文件
# 搭建一个可以识别猫的神经网络
import matplotlib.pyplot as plt
import numpy as np
from Deep_Learning.lr_utils import load_dataset # 加载资料包里面的数据的简单功能的库
# 把数据加载到主程序中
train_set_x_orig, train_set_y, test_set_x_orig, test_set_y, classes = load_dataset()
# 查看训练集里的第26张图片
index = 15
plt.imshow(train_set_x_orig[index])
plt.waitforbuttonpress()
# 查看训练集里的标签
print("train_set_y" + str(train_set_y))
# 打印当前训练集的标签值
# 使用np.squeeze的目的是压缩维度,【未压缩】train_set_y[:,index]的值为[1],【压缩后】np.squeeze(train_set_y[:,index])的值为[1]
print("使用np.squeeze" + str(np.squeeze(train_set_y[:, index])) + ",不使用np.squeeze" + str(train_set_y[:, index]))
print("y=" + str(train_set_y[:, index]) + ",it`s a " + classes[np.squeeze(train_set_y[:, index])].decode("utf-8") + "' picture")
m_train = train_set_y.shape[1] # 训练集里图片数量
m_test = test_set_y.shape[1] # 测试集里图片的数量
# train_set_x_orig是一个维度为(m_train,num_px,num_px,3)的数组
num_px = train_set_x_orig.shape[1] # 训练、测试集里的图片宽度和高度(均为64×64)
# 查看加载的具体情况
print("训练集的数量:m_train = " + str(m_train))
print("测试集的数量:m_test = " + str(m_test))
print("每张图片的宽/高:num_px = " + str(num_px))
print("每张图片的大小:(" + str(num_px) + "," + str(num_px) + ",3)")
print("训练集_图片的维度:" + str(train_set_x_orig.shape))
print("训练集_标签的维度: " + str(train_set_y.shape))
print("测试集_图片的维度:" + str(test_set_x_orig.shape))
print("测试集_标签的维度: " + str(test_set_y.shape))
# 为了方便,把维度为(64,64,3)的numpy数组重构为(64×64×3,1)的数组,若想将形状(a,b,c,d)的矩阵平铺为形状(b*c*d,a)的矩阵可使用以下代码
# 将训练集的维度降低并转置
train_set_x_flatten = train_set_x_orig.reshape(train_set_x_orig.shape[0], -1).T # 其中-1是指不确定是哪一列,若参数为1则为第一列
# 将测试集的围兜降低并转置
test_set_x_flatten = test_set_x_orig.reshape(test_set_x_orig.shape[0], -1).T
# 查看降维之后的具体情况
print("训练集降维之后的维度:" + str(train_set_x_flatten.shape))
print("测试集降维之后的维度:" + str(test_set_x_flatten.shape))
# 标准化数据集,数据集每一行除以255(像素通道的最大值),让标准化的数据位于[0,1]之间
train_set_x = train_set_x_flatten / 255
test_set_x = test_set_x_flatten / 255
# 构建sigmoid函数
def sigmoid(z):
return 1 / (1 + np.exp(-z))
# 测试sigmoid()
print("=========测试sigmoid=========")
print("sigmoid(0) = " + str(sigmoid(0)))
# 初始化参数w和b
def initialize_with_zeros(dim):
"""
此函数为w创建一个维度为(dim,1)的0向量,并将b初始化为0
:param dim: -我们想要的w矢量的大小(或者这种情况下的参数数量)
:return: w -维度为(dim,1)的初始化向量
b -初始化的标量(对应于偏差)
"""
w = np.zeros(shape=(dim, 1))
b = 0
# 使用断言确保数据的正确性
assert (w.shape == (dim, 1)) # w的维度是(dim,1)
assert (isinstance(b, float) or isinstance(b, int)) # b的类型是float或者int
return (w, b)
def propagate(w, b, X, Y):
"""
实现前向和后向传播的成本函数及其梯度
:param w: -权重,大小不等的数组(num_px * num_px * 3,1)
:param b: -偏差,一个标量
:param X: -矩阵类型为(num_px * num_px * 3,训练数量)
:param Y: -真正的"标签"矢量(如果非猫则为0,是猫则是1),矩阵的维度为(1,训练数据数量)
:return:cost -逻辑回归的负对数似然成本
dw -相对于w的损失梯度,因此与w是相同的形状
db -相对于b的损失梯度,因此与b是相同的形状
"""
m = X.shape[1]
# 正向传播
A = sigmoid(np.dot(w.T, X) + b) # 计算激活值
cost = (-1 / m) * np.sum(Y * np.log(A) + (1 - Y) * (np.log(1 - A))) # 计算成本
# 反向传播
dw = (1 / m) * np.dot(X, (A - Y).T)
db = (1 / m) * np.sum(A - Y)
# 使用断言确保数据的正确性
assert (dw.shape == w.shape)
assert (db.dtype == float)
cost = np.squeeze(cost)
assert (cost.shape == ())
# 创建一个字典,保存dw和db
grads = {
"dw": dw,
"db": db
}
return (grads, cost)
# 测试propagate
print("=========测试propagate=========")
# 初始化参数
w, b, X, Y = np.array([[1], [2]]), 2, np.array([[1, 2], [3, 4]]), np.array([[1, 0]])
grads, cost = propagate(w, b, X, Y)
print("dw = " + str(grads["dw"]))
print("db = " + str(grads["db"]))
print("cost = " + str(cost))
# 更新参数,通过最小化成本函数J学习w和b
def optimize(w, b, X, Y, num_iterations, learning_rate, print_cost=False):
"""
通过运行梯度下降算法优化w和b
:param w: -权重,大小不等的数组(num_px * num_px * 3,1 )
:param b: -偏差,标量
:param X: -矩阵类型为(num_px * num_px * 3,训练数量)
:param Y: -真正的"标签"矢量(如果非猫则为0,是猫则是1),矩阵的维度为(1,训练数据数量)
:param num_iterations: -优化循环的迭代次数
:param learning_rate: -梯度下降更新规则的学习速率
:param print_cost: -每100次打印一次损失值
:return:params -包含权重w和b的字典
grads -包含权重和偏差对于成本函数的梯度的字典
cost -优化期间计算的所有成本列表,用于绘制学习曲线
我们需要写下两个步骤遍历:
1.计算当前函数的成本和梯度,使用propagate()
2.使用w和b的梯度下降法更新参数
"""
costs = []
for i in range(num_iterations):
grads, cost = propagate(w, b, X, Y)
dw = grads["dw"]
db = grads["db"]
w = w - learning_rate * dw
b = b - learning_rate * db
# 记录成本
if i % 100 == 0:
costs.append(cost)
# 打印成本数据
if(print_cost) and (i % 100 == 0):
print("迭代的次数:%i, 误差值: %f" % (i, cost))
params = {
"w": w,
"b": b
}
grads = {
"dw": dw,
"db": db
}
return (params, grads, costs)
# 测试optimize
print("=================测试optimize===============")
w, b, X, Y = np.array([[1], [2]]), 2, np.array([[1, 2], [3, 4]]), np.array([[1, 0]])
params, grads, costs = optimize(w, b, X, Y, num_iterations=200, learning_rate=0.009, print_cost=False)
print("w = " + str(params["w"]))
print("b = " + str(params["b"]))
print("dw = " + str(grads["dw"]))
print("db = " + str(grads["db"]))
# 实现预测函数
def predict(w, b, X):
"""
使用学习逻辑回归参数logistic (w,b)预测标签是0还是1
:param w: -权重,大小不等的数组(num_px * num_px * 3,1 )
:param b: -偏差,标量
:param X: -矩阵类型为(num_px * num_px * 3,训练数量)
:return: Y_prediction -包含X中所有图片的所有预测【0 | 1】的一个numpy数组(向量)
"""
m = X.shape[1] # 图片的数量
Y_prediction = np.zeros((1, m))
w = w.reshape(X.shape[0], 1)
# 预测猫在图片中出现的频率
A = sigmoid(np.dot(w.T, X) + b)
for i in range(A.shape[1]):
# 将概率a[0,i]转换为实际预测p[0,i]
Y_prediction[0, 1] = 1 if A[0, i] > 0.5 else 0
assert (Y_prediction.shape == (1, m))
return Y_prediction
# 测试predict
print("============测试predict===============")
w, b, X, Y = np.array([[1], [2]]), 2, np.array([[1, 2], [3, 4]]), np.array([[1, 0]])
print("prediction = " + str(predict(w, b, X)))
# 将函数整合到model函数中
def model(X_train, Y_train, X_test, Y_test, num_iterations=2000, learning_rate=0.5, print_cost=False):
"""
通过调用之前实现的函数构建逻辑回归模型
:param X_train: -numpy的数组,维度为(num_px * num_px * 3,m_train)的训练集
:param Y_train: -numpy的数组,维度为(1,m_train)(矢量)的训练标签集
:param X_test: -numpy的数组,维度为(num_px * num_px * 3,m_train)的测试集
:param Y_test: -numpy的数组,维度为(1,m_train)(矢量)的测试标签集
:param num_iterations: -表示用于优化参数的迭代次数的超参数
:param learning_rate: -表示optimize()更新规则中使用的学习速率的超参数
:param print_cost: -设置为true以每100次的迭代打印成本
:return: d -包含有关模型信息的字典
"""
w, b = initialize_with_zeros(X_train.shape[0])
parameters, grads, costs = optimize(w, b, X_train, Y_train, num_iterations, learning_rate, print_cost)
# 从字典“参数”中检索参数w和b
w, b = parameters["w"], parameters["b"]
# 预测测试/训练集的例子
Y_prediction_train = predict(w, b, X_train)
Y_prediction_test = predict(w, b, X_test)
# 打印训练后的准确性
print("训练集的准确性:", format(100 - np.mean(np.abs(Y_prediction_train - Y_train)) * 100), "%")
print("测试集的准确性:", format(100 - np.mean(np.abs(Y_prediction_test - Y_test)) * 100), "%")
d = {
"costs": costs,
"Y_prediction_train": Y_prediction_train,
"Y_prediction_test": Y_prediction_test,
"w": w,
"b": b,
"learning_rate": learning_rate,
"num_iterations": num_iterations
}
return d
# 测试model
print("===================测试model=====================")
# 这里加载的是真实的数据
d = model(train_set_x, train_set_y, test_set_x, test_set_y, num_iterations=2000, learning_rate=0.005, print_cost=True)
# 绘制图
costs = np.squeeze(d['costs'])
plt.plot(costs)
plt.ylabel('cost')
plt.xlabel('iterations (per hundreds)')
plt.title("learning rate =" + str(d["learning_rate"]))
plt.show()
# 比较模型的学习速率和几种学习速率的选择,也可以尝试使用不同于我们初始化的learning_rate变量包含的三个值
learning_rates = [0.01, 0.001, 0.0001]
models = {}
for i in learning_rates:
print("learning rate is : " + str(i))
models[str(i)] = model(train_set_x, train_set_y, test_set_x, test_set_y, num_iterations=1500, learning_rate=i, print_cost=False)
print("\n" + "---------------------------------" + "\n")
for i in learning_rates:
plt.plot(np.squeeze(models[str(i)]["costs"]), label=str(models[str(i)]["learning_rate"]))
plt.ylabel("cost")
plt.xlabel("iterations")
legend = plt.legend(loc='upper center', shadow=True)
frame = legend.get_frame()
frame.set_facecolor('0.90')
plt.show()
运行结果
显示第15张图片,这不是猫
当学习速率不同时,迭代次数与代价的关系:
控制台运行结果
train_set_y[[0 0 1 0 0 0 0 1 0 0 0 1 0 1 1 0 0 0 0 1 0 0 0 0 1 1 0 1 0 1 0 0 0 0 0 0
0 0 1 0 0 1 1 0 0 0 0 1 0 0 1 0 0 0 1 0 1 1 0 1 1 1 0 0 0 0 0 0 1 0 0 1
0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 1 1 1 0 0 1 0 0 0 0 1 0 1 0 1 1
1 1 1 1 0 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 1 1 0 0 0 1 1 1 1 1 0 0 0 0 1 0
1 1 1 0 1 1 0 0 0 1 0 0 1 0 0 0 0 0 1 0 1 0 1 0 0 1 1 1 0 0 1 1 0 1 0 1
0 0 0 0 0 1 0 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0]]
使用np.squeeze0,不使用np.squeeze[0]
y=[0],it`s a non-cat' picture
训练集的数量:m_train = 209
测试集的数量:m_test = 50
每张图片的宽/高:num_px = 64
每张图片的大小:(64,64,3)
训练集_图片的维度:(209, 64, 64, 3)
训练集_标签的维度: (1, 209)
测试集_图片的维度:(50, 64, 64, 3)
测试集_标签的维度: (1, 50)
训练集降维之后的维度:(12288, 209)
测试集降维之后的维度:(12288, 50)
=========测试sigmoid=========
sigmoid(0) = 0.5
=========测试propagate=========
dw = [[0.99993216]
[1.99980262]]
db = 0.49993523062470574
cost = 6.000064773192205
=================测试optimize===============
w = [[-0.25752876]
[-0.35717757]]
b = 1.4501755960194689
dw = [[0.14547118]
[0.05636579]]
db = -0.04455269818367541
============测试predict===============
prediction = [[0. 1.]]
===================测试model=====================
迭代的次数:0, 误差值: 0.693147
迭代的次数:100, 误差值: 0.584508
迭代的次数:200, 误差值: 0.466949
迭代的次数:300, 误差值: 0.376007
迭代的次数:400, 误差值: 0.331463
迭代的次数:500, 误差值: 0.303273
迭代的次数:600, 误差值: 0.279880
迭代的次数:700, 误差值: 0.260042
迭代的次数:800, 误差值: 0.242941
迭代的次数:900, 误差值: 0.228004
迭代的次数:1000, 误差值: 0.214820
迭代的次数:1100, 误差值: 0.203078
迭代的次数:1200, 误差值: 0.192544
迭代的次数:1300, 误差值: 0.183033
迭代的次数:1400, 误差值: 0.174399
迭代的次数:1500, 误差值: 0.166521
迭代的次数:1600, 误差值: 0.159305
迭代的次数:1700, 误差值: 0.152667
迭代的次数:1800, 误差值: 0.146542
迭代的次数:1900, 误差值: 0.140872
训练集的准确性: 65.55023923444976 %
测试集的准确性: 34.0 %
learning rate is : 0.01
训练集的准确性: 65.55023923444976 %
测试集的准确性: 34.0 %
---------------------------------
learning rate is : 0.001
训练集的准确性: 65.55023923444976 %
测试集的准确性: 34.0 %
---------------------------------
learning rate is : 0.0001
训练集的准确性: 65.55023923444976 %
测试集的准确性: 34.0 %
---------------------------------