Tensorflow--预测乳腺癌

Tensorflow–预测乳腺癌

基于从图像中抽取的反应细胞核的特征数据,采用神经网络算法,借助Tensorflow工具来预测癌症是良性还是恶性.数据是从乳房块的细针抽吸(FNA)的数字化图像数据,它们描述了图像中存在的细胞核的特征.具体步骤如下：

  • 数据说明
  • 数据预处理
  • 数据探索
  • 构建神经网络
  • 训练神经网络
  • 评估模型

一.数据说明

数据可以通过以下方式获取到：链接

从每幅医疗诊断图像中,计算出反应细胞核每个特征的平均值,标准误差,"最差"或最大值(三个最大值的平均值),从而产生30个特征.所有特征值都用四个有效数字重新编码：缺少属性值为None,分类为357良性,212恶性

属性信息：

  • 身份证号码
  • 诊断(M=恶性,B=良性)

计算每个细胞核的10个实值特征：

  • 半径
  • 文理
  • 周长
  • 面积
  • 平滑度
  • 紧凑度
  • 凹面
  • 凹点
  • 对称性
  • 分形维数

查看文件前2行数据：

!head -2 data.csv

"id","diagnosis","radius_mean","texture_mean","perimeter_mean","area_mean","smoothness_mean","compactness_mean","concavity_mean","concave points_mean","symmetry_mean","fractal_dimension_mean","radius_se","texture_se","perimeter_se","area_se","smoothness_se","compactness_se","concavity_se","concave points_se","symmetry_se","fractal_dimension_se","radius_worst","texture_worst","perimeter_worst","area_worst","smoothness_worst","compactness_worst","concavity_worst","concave points_worst","symmetry_worst","fractal_dimension_worst",
842302,M,17.99,10.38,122.8,1001,0.1184,0.2776,0.3001,0.1471,0.2419,0.07871,1.095,0.9053,8.589,153.4,0.006399,0.04904,0.05373,0.01587,0.03003,0.006193,25.38,17.33,184.6,2019,0.1622,0.6656,0.7119,0.2654,0.4601,0.1189

二.数据预处理

1.导入需要的包

import tensorflow as tf
import pandas as pd
from sklearn.utils import shuffle
import matplotlib.gridspec as gridspec
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
import matplotlib.font_manager as fm
myfont=fm.FontProperties(fname="E:/Anaconda/envs/mytensorflow/Lib/site-packages/matplotlib/mpl-data/fonts/ttf/Simhei.ttf")

2.定义数据文件路径：

train_finename="./data/BreastCancer/data.csv"

3.重新设置字段名称

idKey = "id"
diagnosisKey = "diagnosis"
radiusMeanKey = "radius_mean"
textureMeanKey = "texture_mean"
perimeterMeanKey = "perimeter_mean"
areaMeanKey = "area_mean"
smoothnessMeanKey = "smoothness_mean"
compactnessMeanKey = "compactness_mean"
concavityMeanKey = "concavity_mean"
concavePointsMeanKey = "concave points_mean"
symmetryMeanKey = "symmetry_mean"
fractalDimensionMean = "fractal_dimension_mean"
radiusSeKey = "radius_se"
textureSeKey = "texture_se"
perimeterSeKey = "perimeter_se"
areaSeKey = "area_se"
smoothnessSeKey = "smoothness_se"
compactnessSeKey = "compactness_se"
concavitySeKey = "concavity_se"
concavePointsSeKey = "concave points_se"
symmetrySeKey = "symmetry_se"
fractalDimensionSeKey = "fractal_dimension_se"
radiusWorstKey = "radius_worst"
textureWorstKey = "texture_worst"
perimeterWorstKey = "perimeter_worst"
areaWorstKey = "area_worst"
smoothnessWorstKey = "smoothness_worst"
compactnessWorstKey = "compactness_worst"
concavityWorstKey = "concavity_worst"
concavePointsWorstKey = "concave points_worst"
symmetryWorstKey = "symmetry_worst"
fractalDimensionWorstKey = "fractal_dimension_worst"

4.选择训练集列名

train_columns = [idKey, 
                 diagnosisKey, 
                 radiusMeanKey, 
                 textureMeanKey, 
                 perimeterMeanKey, 
                 areaMeanKey, 
                 smoothnessMeanKey, 
                 compactnessMeanKey, 
                 concavityMeanKey, 
                 concavePointsMeanKey, 
                 symmetryMeanKey, 
                 fractalDimensionMean, 
                 radiusSeKey, 
                 textureSeKey, 
                 perimeterSeKey, 
                 areaSeKey, 
                 smoothnessSeKey, 
                 compactnessSeKey, 
                 concavitySeKey, 
                 concavePointsSeKey, 
                 symmetrySeKey, 
                 fractalDimensionSeKey, 
                 radiusWorstKey, 
                 textureWorstKey, 
                 perimeterWorstKey, 
                 areaWorstKey, 
                 smoothnessWorstKey, 
                 compactnessWorstKey, 
                 concavityWorstKey, 
                 concavePointsWorstKey, 
                 symmetryWorstKey, 
                 fractalDimensionWorstKey]

5.定义读取数据函数

文件以逗号分隔,第一行为标题(跳过)：

def get_train_data():
    df=pd.read_csv(train_finename,names=train_columns,delimiter=",",skiprows=1)
    return df

train_data=get_train_data()

三.探索数据

1.查看前5行数据：

train_data.head()
id diagnosis radius_mean texture_mean perimeter_mean area_mean smoothness_mean compactness_mean concavity_mean concave points_mean symmetry_mean fractal_dimension_mean radius_se texture_se perimeter_se area_se smoothness_se compactness_se concavity_se concave points_se symmetry_se fractal_dimension_se radius_worst texture_worst perimeter_worst area_worst smoothness_worst compactness_worst concavity_worst concave points_worst symmetry_worst fractal_dimension_worst
0 842302 M 17.99 10.38 122.80 1001.0 0.11840 0.27760 0.3001 0.14710 0.2419 0.07871 1.0950 0.9053 8.589 153.40 0.006399 0.04904 0.05373 0.01587 0.03003 0.006193 25.38 17.33 184.60 2019.0 0.1622 0.6656 0.7119 0.2654 0.4601 0.11890
1 842517 M 20.57 17.77 132.90 1326.0 0.08474 0.07864 0.0869 0.07017 0.1812 0.05667 0.5435 0.7339 3.398 74.08 0.005225 0.01308 0.01860 0.01340 0.01389 0.003532 24.99 23.41 158.80 1956.0 0.1238 0.1866 0.2416 0.1860 0.2750 0.08902
2 84300903 M 19.69 21.25 130.00 1203.0 0.10960 0.15990 0.1974 0.12790 0.2069 0.05999 0.7456 0.7869 4.585 94.03 0.006150 0.04006 0.03832 0.02058 0.02250 0.004571 23.57 25.53 152.50 1709.0 0.1444 0.4245 0.4504 0.2430 0.3613 0.08758
3 84348301 M 11.42 20.38 77.58 386.1 0.14250 0.28390 0.2414 0.10520 0.2597 0.09744 0.4956 1.1560 3.445 27.23 0.009110 0.07458 0.05661 0.01867 0.05963 0.009208 14.91 26.50 98.87 567.7 0.2098 0.8663 0.6869 0.2575 0.6638 0.17300
4 84358402 M 20.29 14.34 135.10 1297.0 0.10030 0.13280 0.1980 0.10430 0.1809 0.05883 0.7572 0.7813 5.438 94.44 0.011490 0.02461 0.05688 0.01885 0.01756 0.005115 22.54 16.67 152.20 1575.0 0.1374 0.2050 0.4000 0.1625 0.2364 0.07678

2.查看数据特征信息

# 查看数据的统计信息
train_data.describe()
id radius_mean texture_mean perimeter_mean area_mean smoothness_mean compactness_mean concavity_mean concave points_mean symmetry_mean fractal_dimension_mean radius_se texture_se perimeter_se area_se smoothness_se compactness_se concavity_se concave points_se symmetry_se fractal_dimension_se radius_worst texture_worst perimeter_worst area_worst smoothness_worst compactness_worst concavity_worst concave points_worst symmetry_worst fractal_dimension_worst
count 5.690000e+02 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000 569.000000
mean 3.037183e+07 14.127292 19.289649 91.969033 654.889104 0.096360 0.104341 0.088799 0.048919 0.181162 0.062798 0.405172 1.216853 2.866059 40.337079 0.007041 0.025478 0.031894 0.011796 0.020542 0.003795 16.269190 25.677223 107.261213 880.583128 0.132369 0.254265 0.272188 0.114606 0.290076 0.083946
std 1.250206e+08 3.524049 4.301036 24.298981 351.914129 0.014064 0.052813 0.079720 0.038803 0.027414 0.007060 0.277313 0.551648 2.021855 45.491006 0.003003 0.017908 0.030186 0.006170 0.008266 0.002646 4.833242 6.146258 33.602542 569.356993 0.022832 0.157336 0.208624 0.065732 0.061867 0.018061
min 8.670000e+03 6.981000 9.710000 43.790000 143.500000 0.052630 0.019380 0.000000 0.000000 0.106000 0.049960 0.111500 0.360200 0.757000 6.802000 0.001713 0.002252 0.000000 0.000000 0.007882 0.000895 7.930000 12.020000 50.410000 185.200000 0.071170 0.027290 0.000000 0.000000 0.156500 0.055040
25% 8.692180e+05 11.700000 16.170000 75.170000 420.300000 0.086370 0.064920 0.029560 0.020310 0.161900 0.057700 0.232400 0.833900 1.606000 17.850000 0.005169 0.013080 0.015090 0.007638 0.015160 0.002248 13.010000 21.080000 84.110000 515.300000 0.116600 0.147200 0.114500 0.064930 0.250400 0.071460
50% 9.060240e+05 13.370000 18.840000 86.240000 551.100000 0.095870 0.092630 0.061540 0.033500 0.179200 0.061540 0.324200 1.108000 2.287000 24.530000 0.006380 0.020450 0.025890 0.010930 0.018730 0.003187 14.970000 25.410000 97.660000 686.500000 0.131300 0.211900 0.226700 0.099930 0.282200 0.080040
75% 8.813129e+06 15.780000 21.800000 104.100000 782.700000 0.105300 0.130400 0.130700 0.074000 0.195700 0.066120 0.478900 1.474000 3.357000 45.190000 0.008146 0.032450 0.042050 0.014710 0.023480 0.004558 18.790000 29.720000 125.400000 1084.000000 0.146000 0.339100 0.382900 0.161400 0.317900 0.092080
max 9.113205e+08 28.110000 39.280000 188.500000 2501.000000 0.163400 0.345400 0.426800 0.201200 0.304000 0.097440 2.873000 4.885000 21.980000 542.200000 0.031130 0.135400 0.396000 0.052790 0.078950 0.029840 36.040000 49.540000 251.200000 4254.000000 0.222600 1.058000 1.252000 0.291000 0.663800 0.207500 
# 查看是否有空值
train_data.isnull().sum()

id                         0
diagnosis                  0
radius_mean                0
texture_mean               0
perimeter_mean             0
area_mean                  0
smoothness_mean            0
compactness_mean           0
concavity_mean             0
concave points_mean        0
symmetry_mean              0
fractal_dimension_mean     0
radius_se                  0
texture_se                 0
perimeter_se               0
area_se                    0
smoothness_se              0
compactness_se             0
concavity_se               0
concave points_se          0
symmetry_se                0
fractal_dimension_se       0
radius_worst               0
texture_worst              0
perimeter_worst            0
area_worst                 0
smoothness_worst           0
compactness_worst          0
concavity_worst            0
concave points_worst       0
symmetry_worst             0
fractal_dimension_worst    0
dtype: int64

# 查看属于恶性的统计数据
print("恶性")
print(train_data.area_mean[train_data.diagnosis=="M"].describe())

恶性
count     212.000000
mean      978.376415
std       367.937978
min       361.600000
25%       705.300000
50%       932.000000
75%      1203.750000
max      2501.000000
Name: area_mean, dtype: float64

# 查看属于良性的统计数据
print("良性")
print(train_data.area_mean[train_data.diagnosis=="B"].describe())

良性
count    357.000000
mean     462.790196
std      134.287118
min      143.500000
25%      378.200000
50%      458.400000
75%      551.100000
max      992.100000
Name: area_mean, dtype: float64

# 可视化这些数据
f,(ax1,ax2)=plt.subplots(2,1,sharex=True,figsize=(12,4))

bins=50

ax1.hist(train_data.area_mean[train_data.diagnosis=="M"],bins=bins)
ax1.set_title("恶性",fontproperties=myfont)

ax2.hist(train_data.area_mean[train_data.diagnosis=="B"],bins=bins)
ax2.set_title("良性",fontproperties=myfont)

plt.xlabel("区域平均值",fontproperties=myfont)
plt.ylabel("诊断次数",fontproperties=myfont)
plt.show()

output_27_0.png

"area_mean"特征看起来差别比较大,这会增加其在两种类型诊断中的价值.此外恶性诊断更多是均匀分布的,而良性诊断具有正态分布.当其值超过750时,可以更容易做出恶性诊断

4.查看其它特征的特性

r_data=train_data.drop([idKey,areaMeanKey,areaWorstKey,diagnosisKey],axis=1)
r_features=r_data.columns

# 可视化其他特征分布信息
plt.figure(figsize=(12,28*4))
gs=gridspec.GridSpec(28,1)

for i,cn in enumerate(r_data[r_features]):
    ax=plt.subplot(gs[i])
    sns.distplot(train_data[cn][train_data.diagnosis=="M"],bins=50)
    sns.distplot(train_data[cn][train_data.diagnosis=="B"],bins=50)
    ax.set_xlabel("")
    ax.set_title("特征直方图："+str(cn),fontproperties=myfont)
    
plt.show()

output_30_0.png

5.对一些特征进行转换

# 更新诊断值.1代表恶性,0代表为良性
train_data.loc[train_data.diagnosis=="M","diagnosis"]=1
train_data.loc[train_data.diagnosis=="B","diagnosis"]=0

# 创建良性诊断的新特征
train_data.loc[train_data.diagnosis==0,"benign"]=1
train_data.loc[train_data.diagnosis==1,"benign"]=0

# 把这列数据类型转换为int
train_data["benign"]=train_data.benign.astype(int)

# 把列 diagnosis 重命名为 malignant
train_data=train_data.rename(columns={"diagnosis":"malignant"})

# 212例恶性诊断,357例良性诊断.37.25%的诊断是恶性的
print(train_data.benign.value_counts())
print(train_data.malignant.value_counts())

# 查看前几行数据
pd.set_option("display.max_columns",101)

train_data.head()

1    357
0    212
Name: benign, dtype: int64
0    357
1    212
Name: malignant, dtype: int64
id malignant radius_mean texture_mean perimeter_mean area_mean smoothness_mean compactness_mean concavity_mean concave points_mean symmetry_mean fractal_dimension_mean radius_se texture_se perimeter_se area_se smoothness_se compactness_se concavity_se concave points_se symmetry_se fractal_dimension_se radius_worst texture_worst perimeter_worst area_worst smoothness_worst compactness_worst concavity_worst concave points_worst symmetry_worst fractal_dimension_worst benign
0 842302 1 17.99 10.38 122.80 1001.0 0.11840 0.27760 0.3001 0.14710 0.2419 0.07871 1.0950 0.9053 8.589 153.40 0.006399 0.04904 0.05373 0.01587 0.03003 0.006193 25.38 17.33 184.60 2019.0 0.1622 0.6656 0.7119 0.2654 0.4601 0.11890 0
1 842517 1 20.57 17.77 132.90 1326.0 0.08474 0.07864 0.0869 0.07017 0.1812 0.05667 0.5435 0.7339 3.398 74.08 0.005225 0.01308 0.01860 0.01340 0.01389 0.003532 24.99 23.41 158.80 1956.0 0.1238 0.1866 0.2416 0.1860 0.2750 0.08902 0
2 84300903 1 19.69 21.25 130.00 1203.0 0.10960 0.15990 0.1974 0.12790 0.2069 0.05999 0.7456 0.7869 4.585 94.03 0.006150 0.04006 0.03832 0.02058 0.02250 0.004571 23.57 25.53 152.50 1709.0 0.1444 0.4245 0.4504 0.2430 0.3613 0.08758 0
3 84348301 1 11.42 20.38 77.58 386.1 0.14250 0.28390 0.2414 0.10520 0.2597 0.09744 0.4956 1.1560 3.445 27.23 0.009110 0.07458 0.05661 0.01867 0.05963 0.009208 14.91 26.50 98.87 567.7 0.2098 0.8663 0.6869 0.2575 0.6638 0.17300 0
4 84358402 1 20.29 14.34 135.10 1297.0 0.10030 0.13280 0.1980 0.10430 0.1809 0.05883 0.7572 0.7813 5.438 94.44 0.011490 0.02461 0.05688 0.01885 0.01756 0.005115 22.54 16.67 152.20 1575.0 0.1374 0.2050 0.4000 0.1625 0.2364 0.07678 0

6.对数据进行一些预处理

# 创建一个只有 malignant,benign的Dataframe
Malignant=train_data[train_data.malignant==1]
Benign=train_data[train_data.benign==1]

# 将train_x设置为恶性诊断的80%
train_x=Malignant.sample(frac=0.8)
count_Malignants=len(train_x)

# 将80%的良性诊断添加到train_x
train_x=pd.concat([train_x,Benign.sample(frac=0.8)],axis=0)

# 使test_x包含不在train_x中的数据
test_x=train_data.loc[~train_data.index.isin(train_x.index)]

# 使用shuffle函数打乱数据
train_x=shuffle(train_x)
test_x=shuffle(test_x)

# 把标签添加到 train_x,test_y
train_y=train_x.malignant
train_y=pd.concat([train_y,train_x.benign],axis=1)

test_y=test_x.malignant
test_y=pd.concat([test_y,test_x.benign],axis=1)

# 删除train_x,test_x中的标签
train_x=train_x.drop(["malignant","benign"],axis=1)
test_x=test_x.drop(["malignant","benign"],axis=1)

# 核查训练集和测试集数据总数
print(len(train_x))
print(len(train_y))
print(len(test_x))
print(len(test_y))

456
456
113
113

# 提取训练集中所有特征名称
features=train_x.columns.values

# 规范化各特征的值
for feature in features:
    mean,std=train_data[feature].mean(),train_data[feature].std()
    train_x.loc[:,feature]=(train_x[feature]-mean)/std
    test_x.loc[:,feature]=(test_x[feature]-mean)/std

四.构建神经网络

构建1个输入层,4个隐含层,1个输出层的神经网络,隐含层的参数初始化满足正态分布

tf.truncated_normal使用方法如下：

tf.truncated_normal(shape,mean=0.0,stddev=1.0,dtype=tf.float32,seed=None,name=None)

从截断的正态分布中输出随机值.生成的值服从具有指定平均值和标准偏差的正态分布,如果生成的值大于平均值2个标准偏差的值则丢弃并重新选择

五.训练并评估模型

训练模型迭代次数由参数training_epochs确定,批次大小由参数batch_size确定

# 设置参数
learning_rate=0.005
training_dropout=0.9
display_step=1
training_epochs=5
batch_size=100
accuracy_history=[]
cost_history=[]
valid_accuracy_history=[]
valid_cost_history=[]

# 获取输入节点数
input_nodes=train_x.shape[1]
# 设置标签类别数
num_labels=2
# 把测试数据划分为验证集和测试集
split=int(len(test_y)/2)

train_size=train_x.shape[0]
n_samples=train_y.shape[0]

input_x=train_x.as_matrix()
input_y=train_y.as_matrix()
input_x_valid=test_x.as_matrix()[:split]
input_y_valid=test_y.as_matrix()[:split]
input_x_test=test_x.as_matrix()[split:]
input_y_test=test_y.as_matrix()[split:]

E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:22: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:23: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:24: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:25: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:26: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
E:\Anaconda\envs\mytensorflow\lib\site-packages\ipykernel_launcher.py:27: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.

# 设置每个隐含层的节点数
def calculate_hidden_nodes(nodes):
    return (((2*nodes)/3)+num_labels)

hidden_nodes1=round(calculate_hidden_nodes(input_nodes))
hidden_nodes2=round(calculate_hidden_nodes(hidden_nodes1))
hidden_nodes3=round(calculate_hidden_nodes(hidden_nodes2))
print(input_nodes,hidden_nodes1,hidden_nodes2,hidden_nodes3)

31 23 17 13

# 设置保存进行dropout操作时保留节点的比例变量
pkeep = tf.placeholder(tf.float32)

# 定义输入层
x = tf.placeholder(tf.float32, [None, input_nodes])

# 定义第一个隐含层layer1,,初始化为截断的正态分布
W1 = tf.Variable(tf.truncated_normal([input_nodes, hidden_nodes1], stddev = 0.15))
b1 = tf.Variable(tf.zeros([hidden_nodes1]))
y1 = tf.nn.relu(tf.matmul(x, W1) + b1)

# 定义第二个隐含层layer2，初始化为截断的正态分布
W2 = tf.Variable(tf.truncated_normal([hidden_nodes1, hidden_nodes2], stddev = 0.15))
b2 = tf.Variable(tf.zeros([hidden_nodes2]))
y2 = tf.nn.relu(tf.matmul(y1, W2) + b2)

# 定义第三个隐含层layer3,初始化为截断的正态分布
W3 = tf.Variable(tf.truncated_normal([hidden_nodes2, hidden_nodes3], stddev = 0.15)) 
b3 = tf.Variable(tf.zeros([hidden_nodes3]))
y3 = tf.nn.relu(tf.matmul(y2, W3) + b3)
y3 = tf.nn.dropout(y3, pkeep)

# 定义第四个隐含层layer4,初始化为截断的正态分布
W4 = tf.Variable(tf.truncated_normal([hidden_nodes3, 2], stddev = 0.15)) 
b4 = tf.Variable(tf.zeros([2]))
y4 = tf.nn.softmax(tf.matmul(y3, W4) + b4)

# 定义输出层
y=y4
y_=tf.placeholder(tf.float32,[None,num_labels])

# 使用交叉熵最小化误差
cost=-tf.reduce_sum(y_*tf.log(y))

# 使用Adam作为优化器
optimizer=tf.train.AdamOptimizer(learning_rate).minimize(cost)

# 测试模型
correct_prediction=tf.equal(tf.argmax(y,1),tf.argmax(y_,1))

# 计算精度
accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

# 初始化变量
init=tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    
    for epoch in range(training_epochs):
        for batch in range(int(n_samples/batch_size)):
            batch_x=input_x[batch*batch_size:(1+batch)*batch_size]
            batch_y=input_y[batch*batch_size:(1+batch)*batch_size]
            
            sess.run([optimizer],feed_dict={x:batch_x,y_:batch_y,pkeep:training_dropout})
        

         # 循环10次打印日志信息
        if (epoch)%display_step==0:
            train_accuracy,newCost=sess.run([accuracy,cost],feed_dict={x:input_x,y_:input_y,pkeep:training_dropout})
            valid_accuracy,valid_newCost=sess.run([accuracy,cost],feed_dict={x:input_x_valid,y_:input_y_valid,pkeep:1})
            print("Epoch:",epoch,"Acc=","{:.5f}".format(train_accuracy),
                      "Cost=","{:.5f}".format(newCost),
                      "Valid_Acc=","{:.5f}".format(valid_accuracy),
                      "Valid_Cost=","{:.5f}".format(valid_newCost))
            # 记录模型结果
            accuracy_history.append(train_accuracy)
            cost_history.append(newCost)
            valid_accuracy_history.append(valid_accuracy)
            valid_cost_history.append(valid_newCost)
                
            # 如若15次日志信息并没有改善,停止迭代
            if valid_accuracy<max(valid_accuracy_history) and epoch >100:
                stop_early+=1
                if stop_early==15:
                    break
            else:
                stop_early=0
                    
    # 可视化精度及损失值
    f,(ax1,ax2)=plt.subplots(2,1,sharex=True,figsize=(10,4))
    
    ax1.plot(accuracy_history,color="b")
    ax1.plot(valid_accuracy_history,color="g")
    ax1.set_title("精度",fontproperties=myfont)
    
    ax2.plot(cost_history,color="b")
    ax2.plot(valid_cost_history,color="g")
    ax2.set_title("损失值",fontproperties=myfont)
    
    plt.xlabel("迭代次数(x10)",fontproperties=myfont)
    plt.show()

Epoch: 0 Acc= 0.71053 Cost= 291.99823 Valid_Acc= 0.83929 Valid_Cost= 35.09937
Epoch: 1 Acc= 0.79386 Cost= 249.30136 Valid_Acc= 0.85714 Valid_Cost= 28.89331
Epoch: 2 Acc= 0.88377 Cost= 194.89590 Valid_Acc= 0.94643 Valid_Cost= 20.11331
Epoch: 3 Acc= 0.93421 Cost= 138.11267 Valid_Acc= 0.96429 Valid_Cost= 11.85844
Epoch: 4 Acc= 0.94518 Cost= 95.28279 Valid_Acc= 0.98214 Valid_Cost= 6.43805

output_46_1.png

随着训练次数的不断增加,模型的精度越来越高,损失值越来越小

posted @ 2019-04-10 20:29  LQ6H  阅读(430)  评论(0)  编辑  收藏  举报