练习题答案及解说

第3章

第一题：三个权重应为接近于0.4、0.4、0.2的数值；例子代码请参看score1q1.py。
第二题：例子代码请参看score1q2.py。运行程序后观看结果，有可能发现权重参数出现异常情况，例如出现负值。

第4章

第一题：例子代码（codeq1.py）如下：

import tensorflow as tf

a = tf.placeholder(shape=[3, 2], dtype=tf.int32)

b = a * 7

sess = tf.Session()

print(sess.run(b, feed_dict={a: [[1, 2], [3, 4], [5, 6]]}))

第二题：例子代码（codeq2.py）如下，注意softmax函数要求参数是浮点数，所以定义张量a的时候改成了tf.float32：

import tensorflow as tf

a = tf.placeholder(shape=[3, 2], dtype=tf.float32)

b = a * 7

c = tf.nn.softmax(b)

sess = tf.Session()

print(sess.run(c, feed_dict={a: [[1, 2], [3, 4], [5, 6]]}))

第5章

第一题：例子代码（codeq3.py）如下：

import tensorflow as tf
import random

a = tf.placeholder(dtype=tf.float32)

b = tf.nn.sigmoid(a)

sess = tf.Session()

for i in range(5):
    print(sess.run(b, feed_dict={a: random.random() * 40 - 20}))

第二题：例子代码（codeq4.py）如下：

import tensorflow as tf

x = tf.placeholder(shape=[3], dtype=tf.float32)
yTrain = tf.placeholder(shape=[], dtype=tf.float32)

w = tf.Variable(tf.zeros([3]), dtype=tf.float32)
b = tf.Variable(0, dtype=tf.float32)

y = tf.reduce_sum(tf.nn.sigmoid(x * w + b))

loss = tf.abs(y - yTrain)

optimizer = tf.train.RMSPropOptimizer(0.1)

train = optimizer.minimize(loss)

sess = tf.Session()

sess.run(tf.global_variables_initializer())

for i in range(1000):
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [1, 1, 1], yTrain: 2}))
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [1, 0, 1], yTrain: 1}))
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [1, 2, 3], yTrain: 3}))

第6章

第一题：例子代码如下：

import numpy as np
import pandas as pd


fileData = pd.read_csv('dataq.csv', dtype=np.float32, header=None, usecols=(0, 1, 3, 4))

wholeData = fileData.as_matrix()

print(wholeData)

第二题：例子代码（codeq5.py）如下：

import numpy as np
import pandas as pd
import random


fileData = pd.read_csv('dataq.csv', dtype=np.float32, header=None, converters={(2): lambda s: random.random()})

wholeData = fileData.as_matrix()

print(wholeData)

第7章

第一题：例子代码（codeq6.py）如下：

import tensorflow as tf
import numpy as np

m1 = tf.placeholder(shape=[3, 3], dtype=tf.float32)
m2 = tf.placeholder(shape=[3, 4], dtype=tf.float32)

y = tf.matmul(m1, m2)

sess = tf.Session()

print(sess.run([y], feed_dict={m1: [[1, 2, 3], [7, 8, 9], [4, 5, 6]], m2: [[1, 2, 3, 4], [3, 4, 5, 6], [5, 6, 7, 8]]}))

print(np.matmul([[1, 2, 3], [7, 8, 9], [4, 5, 6]], [[1, 2, 3, 4], [3, 4, 5, 6], [5, 6, 7, 8]]))

第二题：例子代码（codeq7.py）如下：

import tensorflow as tf

x = tf.placeholder(shape=[1, 3], dtype=tf.float32)
yTrain = tf.placeholder(shape=[], dtype=tf.float32)

w = tf.Variable(tf.zeros([3, 3]), dtype=tf.float32)
b = tf.Variable(0, dtype=tf.float32)

y = tf.reduce_sum(tf.nn.sigmoid(tf.matmul(x, w) + b))

loss = tf.abs(y - yTrain)

optimizer = tf.train.RMSPropOptimizer(0.01)

train = optimizer.minimize(loss)

sess = tf.Session()

sess.run(tf.global_variables_initializer())

for i in range(2000):
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [[1, 1, 1]], yTrain: 2}))
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [[1, 0, 1]], yTrain: 1}))
    print(sess.run([train, y, yTrain, loss], feed_dict={x: [[1, 2, 3]], yTrain: 3}))

第8章

第一题：例子代码如下：

import tensorflow as tf
import sys

argt = sys.argv[1:]

for v in argt:
    if v == "-v":
        print(tf.__version__)

第二题：例子代码如下：

import tensorflow as tf
import random
import os
import sys

ifRestartT = False

argt = sys.argv[1:]

for v in argt:
    if v == "-restart":
        ifRestartT = True


trainResultPath = "./save/idcard2"

random.seed()

x = tf.placeholder(tf.float32)
yTrain = tf.placeholder(tf.float32)

w1 = tf.Variable(tf.random_normal([4, 8], mean=0.5, stddev=0.1), dtype=tf.float32)
b1 = tf.Variable(0, dtype=tf.float32)

xr = tf.reshape(x, [1, 4])

n1 = tf.nn.tanh(tf.matmul(xr, w1)  + b1)

w2 = tf.Variable(tf.random_normal([8, 2], mean=0.5, stddev=0.1), dtype=tf.float32)
b2 = tf.Variable(0, dtype=tf.float32)

n2 = tf.matmul(n1, w2) + b2

y = tf.nn.softmax(tf.reshape(n2, [2]))

loss = tf.reduce_mean(tf.square(y - yTrain))

optimizer = tf.train.RMSPropOptimizer(0.01)

train = optimizer.minimize(loss)

sess = tf.Session()

if ifRestartT == True:
    print("force restart...")
    sess.run(tf.global_variables_initializer())
elif os.path.exists(trainResultPath + ".index"):
    resultT = input('Would you like to load? (y/n)')

    if resultT == "y":
        print("loading: %s" % trainResultPath)
        tf.train.Saver().restore(sess, save_path=trainResultPath)
    else:
        print("initializing...")
        sess.run(tf.global_variables_initializer())
else:
    print("train result path not exists: %s" % trainResultPath)
    sess.run(tf.global_variables_initializer())


lossSum = 0.0

for i in range(5):

    xDataRandom = [int(random.random() * 10), int(random.random() * 10), int(random.random() * 10), int(random.random() * 10)]
    if xDataRandom[2] % 2 == 0:
        yTrainDataRandom = [0, 1]
    else:
        yTrainDataRandom = [1, 0]

    result = sess.run([train, x, yTrain, y, loss], feed_dict={x: xDataRandom, yTrain: yTrainDataRandom})

    lossSum = lossSum + float(result[len(result) - 1])

    print("i: %d, loss: %10.10f, avgLoss: %10.10f" % (i, float(result[len(result) - 1]), lossSum / (i + 1)))

    if os.path.exists("save.txt"):
        os.remove("save.txt")
        print("saving...")
        tf.train.Saver().save(sess, save_path=trainResultPath)

resultT = input('Would you like to save? (y/n)')

if resultT == "y":
    print("saving...")
    tf.train.Saver().save(sess, save_path=trainResultPath)

第9章

第一题：例子代码如下，注意可变参数w的形态要做相应改变：

import tensorflow as tf

x = tf.placeholder(shape=[1, 3], dtype=tf.float32)

w = tf.Variable(tf.ones([1, 3]), dtype=tf.float32)
b = tf.Variable(1, dtype=tf.float32)

y = x * w + b

sess = tf.Session()
sess.run(tf.global_variables_initializer())

sess.run(y, feed_dict={x: [[1, 2, 3]]})

writer = tf.summary.FileWriter("graph", sess.graph)

第二题：例子代码如下，其中用tf.multiply代替点乘的乘法符号即可给操作起名：

import tensorflow as tf

x = tf.placeholder(shape=[1, 3], dtype=tf.float32)

w = tf.Variable(tf.ones([1, 3]), dtype=tf.float32)
b = tf.Variable(1, dtype=tf.float32)

y = tf.multiply(x, w, name="MyMul") + b

sess = tf.Session()
sess.run(tf.global_variables_initializer())

sess.run(y, feed_dict={x: [[1, 2, 3]]})

writer = tf.summary.FileWriter("graph", sess.graph)

第10章

第一题：例子代码（codeq8.py）如下：

import tensorflow as tf
import numpy as np
import sys

predictData = None

argt = sys.argv[1:]

for v in argt:
    if v.startswith("-predict="):
        tmpStr = v[len("-predict="):]
        predictData = np.fromstring(tmpStr, dtype=np.float32, sep=",")

print("predictData: %s" % predictData)

x1 = tf.placeholder(dtype=tf.float32)
x2 = tf.placeholder(dtype=tf.float32)
x3 = tf.placeholder(dtype=tf.float32)
yTrain = tf.placeholder(dtype=tf.float32)

w1 = tf.Variable(0.1, dtype=tf.float32)
w2 = tf.Variable(0.1, dtype=tf.float32)
w3 = tf.Variable(0.1, dtype=tf.float32)

n1 = x1 * w1
n2 = x2 * w2
n3 = x3 * w3

y = n1 + n2 + n3

loss = tf.abs(y - yTrain)

optimizer = tf.train.RMSPropOptimizer(0.001)

train = optimizer.minimize(loss)

sess = tf.Session()

init = tf.global_variables_initializer()

sess.run(init)

for i in range(200):
	result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 90, x2: 80, x3: 70, yTrain: 85})
	print(result)

	result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 98, x2: 95, x3: 87, yTrain: 96})
	print(result)

if predictData is not None:
    print(sess.run(y, feed_dict={x1: predictData[0], x2: predictData[1], x3: predictData[2]}))

第二题：例子代码（codeq9.py）如下，程序最后会要求输入需预测的数据，按类似“70,70,70”的方法输入即可：

import tensorflow as tf
import numpy as np
import sys


x1 = tf.placeholder(dtype=tf.float32)
x2 = tf.placeholder(dtype=tf.float32)
x3 = tf.placeholder(dtype=tf.float32)
yTrain = tf.placeholder(dtype=tf.float32)

w1 = tf.Variable(0.1, dtype=tf.float32)
w2 = tf.Variable(0.1, dtype=tf.float32)
w3 = tf.Variable(0.1, dtype=tf.float32)

n1 = x1 * w1
n2 = x2 * w2
n3 = x3 * w3

y = n1 + n2 + n3

loss = tf.abs(y - yTrain)

optimizer = tf.train.RMSPropOptimizer(0.001)

train = optimizer.minimize(loss)

sess = tf.Session()

init = tf.global_variables_initializer()

sess.run(init)

for i in range(200):
	result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 90, x2: 80, x3: 70, yTrain: 85})
	print(result)

	result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 98, x2: 95, x3: 87, yTrain: 96})
	print(result)

resultT = input('Please input data: ')

predictData = np.fromstring(resultT, dtype=np.float32, sep=",")
print("predictData: %s" % predictData)

if predictData is not None:
    print(sess.run(y, feed_dict={x1: predictData[0], x2: predictData[1], x3: predictData[2]}))

第11章

第一题：例子代码（codeq10.py）如下，其中激活函数“linear”代表不用激活函数，而是线性的：

import tensorflow.contrib.keras as k
import numpy as np


model = k.models.Sequential()

model.add(k.layers.Dense(3, input_dim=3, activation='linear'))

model.add(k.layers.Dense(1, input_dim=3, activation='linear'))

model.compile(loss='mean_squared_error', optimizer="RMSProp", metrics=['accuracy'])

model.fit([[90, 80, 70], [98, 95, 87]], [85, 96], epochs=10000, batch_size=1, verbose=2)

xTestData = np.array([[80, 80, 80], [99, 98, 97]], dtype=np.float32)

resultAry = model.predict(xTestData)
print(resultAry)

第二题：例子代码（codeq11.py）如下：

import tensorflow.contrib.keras as k
import numpy as np


model = k.models.Sequential()

model.add(k.layers.Dense(8, input_dim=3, activation='tanh'))

model.add(k.layers.Dense(1, input_dim=3, activation='linear'))

model.compile(loss='mean_squared_error', optimizer="RMSProp", metrics=['accuracy'])

model.fit([[1, 1, 1], [1, 0, 1], [1, 2, 3]], [2, 1, 3], epochs=1000, batch_size=1, verbose=2)

xTestData = np.array([[1, 2, 2], [2, 3, 3]], dtype=np.float32)

resultAry = model.predict(xTestData)
print(resultAry)

第13章

第一题：例子代码（codeq12.py）如下，请自行准备一个包含灰度像素的文本文件checkData64.txt，其中每行前64个数字是8*8矩阵中每个点的灰度，灰度值为0-255，最后3个数字还是3种情况的概率，即每行67个数字：

import tensorflow as tf
import numpy as np
import pandas as pd
import sys


roundCount = 100
learnRate = 0.01

argt = sys.argv[1:]

for v in argt:
    if v.startswith("-round="):
        roundCount = int(v[len("-round="):])
    if v.startswith("-learnrate="):
        learnRate = float(v[len("-learnrate="):])


fileData = pd.read_csv('checkData64.txt', dtype=np.float32, header=None)

wholeData = fileData.as_matrix()

rowCount = wholeData.shape[0]

print("wholeData=%s" % wholeData)
print("rowSize=%d" % wholeData.shape[1])
print("rowCount=%d" % rowCount)

x = tf.placeholder(shape=[64], dtype=tf.float32)
yTrain = tf.placeholder(shape=[3], dtype=tf.float32)

filter1T = tf.Variable(tf.ones([2, 2, 1, 1]), dtype=tf.float32)

n1 = tf.nn.conv2d(input=tf.reshape(x, [1, 8, 8, 1]), filter=filter1T, strides=[1, 1, 1, 1], padding='SAME')

filter2T = tf.Variable(tf.ones([2, 2, 1, 1]), dtype=tf.float32)

n2 = tf.nn.conv2d(input=tf.reshape(n1, [1, 8, 8, 1]), filter=filter2T, strides=[1, 1, 1, 1], padding='VALID')

filter3T = tf.Variable(tf.ones([2, 2, 1, 1]), dtype=tf.float32)

n3 = tf.nn.conv2d(input=tf.reshape(n2, [1, 7, 7, 1]), filter=filter3T, strides=[1, 1, 1, 1], padding='VALID')

filter4T = tf.Variable(tf.ones([2, 2, 1, 1]), dtype=tf.float32)

n4 = tf.nn.conv2d(input=tf.reshape(n3, [1, 6, 6, 1]), filter=filter4T, strides=[1, 1, 1, 1], padding='VALID')

n4f = tf.reshape(n4, [1, 25])

w4 = tf.Variable(tf.random_normal([25, 32]), dtype=tf.float32)
b4 = tf.Variable(0, dtype=tf.float32)

n4 = tf.nn.tanh(tf.matmul(n4f, w4) + b4)

w5 = tf.Variable(tf.random_normal([32, 3]), dtype=tf.float32)
b5 = tf.Variable(0, dtype=tf.float32)

n5 = tf.reshape(tf.matmul(n4, w5) + b5, [-1])

y = tf.nn.softmax(n5)

loss = -tf.reduce_mean(yTrain * tf.log(tf.clip_by_value(y, 1e-10, 1.0)))
optimizer = tf.train.RMSPropOptimizer(learnRate)

train = optimizer.minimize(loss)

sess = tf.Session()
sess.run(tf.global_variables_initializer())

for i in range(roundCount):
    lossSum = 0.0

    for j in range(rowCount):
        result = sess.run([train, x, yTrain, y, loss], feed_dict={x: wholeData[j][0:64], yTrain: wholeData[j][64:67]})

        lossT = float(result[len(result) - 1])

        lossSum = lossSum + lossT

        if j == (rowCount - 1):
            print("i: %d, loss: %10.10f, avgLoss: %10.10f" % (i, lossT, lossSum / (rowCount + 1)))

第14章

第一题：例子代码（lstmq1.py）如下，请自行准备一个文本文件exchangeData2.txt，其中每行包含两个数字，第一个是当日汇率，第二个是当日交易量，注意在计算y时reduce_sum函数最后一个参数axis是指在哪个维度上进行求和，可以自行输出不加这个参数和加这个参数后y的形态作对比：

import tensorflow as tf
import numpy as np
import pandas as pd
import sys

roundT = 1000
learnRateT = 0.1

argt = sys.argv[1:]
print("argt: %s" % argt)

for v in argt:
    if v.startswith("-round="):
        roundT = int(v[len("-round="):])
    if v.startswith("-learnrate="):
        learnRateT = float(v[len("-learnrate="):])

fileData = pd.read_csv('exchangeData2.txt', dtype=np.float32, header=None)
wholeData = np.reshape(fileData.as_matrix(), (-1, 2))

print("wholeData: %s" % wholeData)

cellCount = 3
unitCount = 5

testData = wholeData[-cellCount:]
print("testData: %s\n" % testData)

rowCount = wholeData.shape[0] - cellCount
print("rowCount: %d\n" % rowCount)

xData = [wholeData[i:i + cellCount] for i in range(rowCount)]
yTrainData = [wholeData[i + cellCount] for i in range(rowCount)]

print("xData: %s\n" % xData)
print("yTrainData: %s\n" % yTrainData)

x = tf.placeholder(shape=[cellCount, 2], dtype=tf.float32)
yTrain = tf.placeholder(dtype=tf.float32)

cellT = tf.nn.rnn_cell.BasicLSTMCell(unitCount)

initState = cellT.zero_state(1, dtype=tf.float32)

h, finalState = tf.nn.dynamic_rnn(cellT, tf.reshape(x, [1, cellCount, 2]), initial_state=initState, dtype=tf.float32)
hr = tf.reshape(h, [cellCount, unitCount])

w2 = tf.Variable(tf.random_normal([unitCount, 2]), dtype=tf.float32)
b2 = tf.Variable(tf.zeros([2]), dtype=tf.float32)

y = tf.reduce_sum(tf.matmul(hr, w2) + b2, axis=0)

loss = tf.reduce_mean(tf.square(y - yTrain))

optimizer = tf.train.RMSPropOptimizer(learnRateT)
train = optimizer.minimize(loss)

sess = tf.Session()
sess.run(tf.global_variables_initializer())

for i in range(roundT):
    lossSum = 0.0
    for j in range(rowCount):
        result = sess.run([train, x, yTrain, y, h, finalState, loss], feed_dict={x: xData[j], yTrain: yTrainData[j]})
        lossSum = lossSum + float(result[len(result) - 1])
        if j == (rowCount - 1):
            print("i: %d, x: %s, yTrain: %s, y: %s, h: %s, finalState: %s, loss: %s, avgLoss: %10.10f\n" % (i, result[1], result[2], result[3], result[4], result[5], result[6], (lossSum / rowCount)))

result = sess.run([x, y], feed_dict={x: testData})
print("x: %s, y: %s\n" % (result[0], result[1]))