TensorFlow手写数字识别

标签：loss val ... 0.4241323 train TensorFlow 手写识别 accuracy

In [1]:

import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from sklearn.preprocessing import StandardScaler
import pandas as pd

In [2]:

(x_train_all, y_train_all), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

In [3]:

x_valid, x_train = x_train_all[:5000], x_train_all[5000:]
y_valid, y_train = y_train_all[:5000], y_train_all[5000:]

In [4]:

scaler = StandardScaler()
x_train.shape

Out[4]:

(55000, 28, 28)

In [5]:

x_train = x_train.astype(np.float32).reshape(-1, 1)
x_train.shape

Out[5]:

(43120000, 1)

In [6]:

x_train_scaled = scaler.fit_transform(x_train)
x_train_scaled

Out[6]:

array([[-0.4241323],
       [-0.4241323],
       [-0.4241323],
       ...,
       [-0.4241323],
       [-0.4241323],
       [-0.4241323]], dtype=float32)

In [7]:

x_train_scaled = x_train_scaled.reshape(-1, 28, 28)
display(x_train_scaled.shape)
x_train_scaled

(55000, 28, 28)

Out[7]:

array([[[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]],

       [[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]],

       [[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]],

       ...,

       [[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]],

       [[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]],

       [[-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        ...,
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323],
        [-0.4241323, -0.4241323, -0.4241323, ..., -0.4241323,
         -0.4241323, -0.4241323]]], dtype=float32)

In [8]:

x_valid_scaled = scaler.fit_transform(x_valid.astype(np.float32).reshape(-1,1)).reshape(-1, 28, 28)
x_test_scaled = scaler.fit_transform(x_test.astype(np.float32).reshape(-1,1)).reshape(-1, 28, 28)

In [9]:

model = tf.keras.models.Sequential([
    tf.keras.layers.Flatten(input_shape=[28, 28]),    # 定义输入数据的格式
    tf.keras.layers.Dense(512, activation='relu', kernel_regularizer='l2'),
    # tf.keras.layers.AlphaDropout(0.5),  也可以取消正则化，使用改进版的dropout（均值和方差不变，归一化性质也不变）
    tf.keras.layers.Dense(256, activation='relu', kernel_regularizer='l2'),
    tf.keras.layers.Dense(128, activation='relu', kernel_regularizer='l2'),
    tf.keras.layers.Dense(10, activation='softmax')
])

# 配置模型
model.compile(loss='sparse_categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])

In [10]:

history = model.fit(x_train_scaled, y_train, epochs=10, validation_data=(x_valid_scaled, y_valid))

Epoch 1/10
1719/1719 [==============================] - 5s 3ms/step - loss: 8.6183 - accuracy: 0.8999 - val_loss: 6.0358 - val_accuracy: 0.9376
Epoch 2/10
1719/1719 [==============================] - 5s 3ms/step - loss: 4.4725 - accuracy: 0.9445 - val_loss: 3.2236 - val_accuracy: 0.9510
Epoch 3/10
1719/1719 [==============================] - 5s 3ms/step - loss: 2.4441 - accuracy: 0.9506 - val_loss: 1.8074 - val_accuracy: 0.9582
Epoch 4/10
1719/1719 [==============================] - 5s 3ms/step - loss: 1.4214 - accuracy: 0.9556 - val_loss: 1.0957 - val_accuracy: 0.9600
Epoch 5/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.9005 - accuracy: 0.9595 - val_loss: 0.7379 - val_accuracy: 0.9586
Epoch 6/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.6334 - accuracy: 0.9615 - val_loss: 0.5429 - val_accuracy: 0.9670
Epoch 7/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.4934 - accuracy: 0.9639 - val_loss: 0.4449 - val_accuracy: 0.9680
Epoch 8/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.4187 - accuracy: 0.9654 - val_loss: 0.3877 - val_accuracy: 0.9688
Epoch 9/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.3773 - accuracy: 0.9665 - val_loss: 0.3654 - val_accuracy: 0.9682
Epoch 10/10
1719/1719 [==============================] - 5s 3ms/step - loss: 0.3529 - accuracy: 0.9679 - val_loss: 0.3481 - val_accuracy: 0.9684

In [11]:

history.history

Out[11]:

{'loss': [8.618253707885742,
  4.4725117683410645,
  2.4440758228302,
  1.4214469194412231,
  0.9005103707313538,
  0.633385956287384,
  0.49344661831855774,
  0.4187164306640625,
  0.3772663474082947,
  0.3529081642627716],
 'accuracy': [0.8999454379081726,
  0.9445454478263855,
  0.9505817890167236,
  0.9556363821029663,
  0.9594908952713013,
  0.961509108543396,
  0.9638727307319641,
  0.9654181599617004,
  0.9665091037750244,
  0.9678727388381958],
 'val_loss': [6.035829067230225,
  3.2235751152038574,
  1.8073712587356567,
  1.0957262516021729,
  0.7379411458969116,
  0.5428724884986877,
  0.4449375569820404,
  0.38767603039741516,
  0.36538875102996826,
  0.34807276725769043],
 'val_accuracy': [0.9376000165939331,
  0.9509999752044678,
  0.9581999778747559,
  0.9599999785423279,
  0.9585999846458435,
  0.9670000076293945,
  0.9679999947547913,
  0.9688000082969666,
  0.9682000279426575,
  0.9684000015258789]}

In [12]:

pd.DataFrame(history.history).plot(figsize=(8,5))
plt.grid()
plt.gca().set_ylim(0,1)
plt.show()

No description has been provided for this image

In [13]:

model.evaluate(x_test_scaled, y_test)

313/313 [==============================] - 0s 1ms/step - loss: 0.3486 - accuracy: 0.9644

Out[13]:

[0.34860044717788696, 0.9643999934196472]

In [14]:

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 flatten (Flatten)           (None, 784)               0                                                                    
 dense (Dense)               (None, 512)               401920                                                         
 dense_1 (Dense)             (None, 256)               131328                                                       
 dense_2 (Dense)             (None, 128)               32896                                                           
 dense_3 (Dense)             (None, 10)                1290                                                               
=================================================================
Total params: 567434 (2.16 MB)
Trainable params: 567434 (2.16 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

In [ ]:

标签：loss,val,...,0.4241323,train,TensorFlow,手写,识别,accuracy
From： https://www.cnblogs.com/funsion/p/18679462

相关文章

赞助商

阅读排行