目錄
0 前言
比賽網址：https://www.kaggle.com/c/digit-recognizer
參考解法：https://www.kaggle.com/yassineghouzam/introduction-to-cnn-keras-0-997-top-6
需要用到的庫：pandas、numpy、matplotlib、seaborn、sklearn、keras
1 簡介
- 卷機網絡模型
- 基於Keras
- 准確率99.6%
- 1080Ti，30個epoch，訓練時間3min
導包：
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import seaborn as sns
# %matplotlib inline
np.random.seed(2)
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
import itertools
from keras.utils.np_utils import to_categorical # convert to one-hot-encoding
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D
from keras.optimizers import RMSprop
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ReduceLROnPlateau
sns.set(style='white', context='notebook', palette='deep')
2 數據准備
2.1 導入數據
# Load the data
train = pd.read_csv("../input/train.csv")
test = pd.read_csv("../input/test.csv")
Y_train = train["label"]
X_train = train.drop(labels = ["label"],axis = 1)
# free some space
del train
g = sns.countplot(Y_train)
Y_train.value_counts()
1 4684
7 4401
3 4351
9 4188
2 4177
6 4137
0 4132
4 4072
8 4063
5 3795
Name: label, dtype: int64
2.2 檢查空值
X_train.isnull().any().describe()
test.isnull().any().describe()
X_train
count 784
unique 1
top False
freq 784
dtype: object
--------------------
test
count 784
unique 1
top False
freq 784
dtype: object
X_train.isnull()會將整個數據表中所有項轉換為bool型，空值為TrueX_train.isnull().any()會將所有列進行統計，列中包含空值，該列為True。可指定參數axis=1，對行進行統計X_train.isnull().any().describe()將信息進行匯總。count為總統計數；unique為種類（由於該數據中沒有空值，全為False，故只有1類）；top為最多的種類；freq為最多種類出現頻次
2.3 正則化 Normalization
將數據正則化到[0, 1]范圍內，減小光照的影響，並可加速CNN收斂速度
X_train = X_train / 255.0
test = test / 255.0
2.4 更改數據維度 Reshape
X_train = X_train.values.reshape(-1,28,28,1)
test = test.values.reshape(-1,28,28,1)
- 784個像素，28x28=784
- 最后需要多加一維，代表通道數（在Keras中需要）。
2.5 標簽編碼
將標簽轉化為one-hot類型
# eg. 2 -> [0,0,1,0,0,0,0,0,0,0])
Y_train = to_categorical(Y_train, num_classes = 10)
2.6 分割交叉驗證集
random_seed = 2 #可去掉
X_train, X_val, Y_train, Y_val = train_test_split(X_train, Y_train, test_size = 0.1, random_state=random_seed)
train_test_split()函數說明：
Signature: train_test_split(*arrays, **options) Docstring: Split arrays or matrices into random train and test subsets Parameters ---------- *arrays : sequence of indexables with same length / shape[0] Allowed inputs are lists, numpy arrays, scipy-sparse matrices or pandas dataframes. test_size : float, int or None, optional (default=0.25). 分出來測試集的百分比 train_size : float, int, or None, (default=None). 訓練集百分比，和上面那個寫一個就行，另外一個會自動算 random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by `np.random`. shuffle : boolean, optional (default=True) Whether or not to shuffle the data before splitting. If shuffle=False then stratify must be None. stratify : array-like or None (default=None) If not None, data is split in a stratified fashion, using this as the class labels. 一些不平衡的數據集需要添加這個參數
# Some examples
g = plt.imshow(X_train[0][:,:,0])
3 CNN
3.1 定義網絡模型
網絡結構：
- First Layer
- Conv2D 5x5x32
- Conv2D 5x5x32
- maxpool, size 2,2
- Second Layer
- Conv2D 3x3x64
- Conv2D 3x3x64
- maxpool, size 2,2 strides 2,2
- Third Layer
- Flatten
- Forth Layer
- full connect 256
- Fifth Layer
- full connect 10
model = Sequential()
model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same',
activation ='relu', input_shape = (28,28,1)))
model.add(Conv2D(filters = 32, kernel_size = (5,5),padding = 'Same',
activation ='relu'))
model.add(MaxPool2D(pool_size=(2,2)))
#dropout用於正則化項，隨機丟失一些節點，防止網絡過擬合
model.add(Dropout(0.25))
model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same',
activation ='relu'))
model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same',
activation ='relu'))
model.add(MaxPool2D(pool_size=(2,2), strides=(2,2)))
model.add(Dropout(0.25))
# 將特征圖轉換成1D向量
model.add(Flatten())
model.add(Dense(256, activation = "relu"))
model.add(Dropout(0.5))
model.add(Dense(10, activation = "softmax"))
3.2 設置優化器和退火器 optimizer and annealer
定義好網絡后，需設置得分函數、損失函數和優化算法。
- 優化器
- 這里選用RMS，還可以用Adam等
- 深度學習優化入門：Momentum、RMSProp 和 Adam
optimizer = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)
- 編譯模型
- 參數
metrics=["accuracy"]用於評估模型的表現
- 參數
model.compile(optimizer = optimizer , loss = "categorical_crossentropy", metrics=["accuracy"])
- 對學習率（LR）使用退火算法（annealing method）？
- 為了更快的接近全局最小值，Loss大時LR大，之后逐步減小LR
- 當准確率X步后不再提高，LR減半
learning_rate_reduction = ReduceLROnPlateau(monitor='val_acc',
patience=3,
verbose=1,
factor=0.5,
min_lr=0.00001)
- epochs和batch_size
- 過完一次完整的數據集是一個epoch
- 當數據集過大時，將數據集分成多個batch，依次訓練
epochs = 1 # Turn epochs to 30 to get 0.9967 accuracy
batch_size = 86
3.3 數據增強
為了防止過擬合問題，使用一些方法擴大我們的數據集。
最理想的是使用一些小的變化改變訓練數據。如：旋轉、翻轉、隨機剪裁、縮放等
使用效果：
- 數據增強前，准確率98.1%
- 數據增強后，准確率99.6%
# 使用數據增強算法
datagen = ImageDataGenerator(
featurewise_center=False, # set input mean to 0 over the dataset
samplewise_center=False, # set each sample mean to 0
featurewise_std_normalization=False, # divide inputs by std of the dataset
samplewise_std_normalization=False, # divide each input by its std
zca_whitening=False, # apply ZCA whitening
rotation_range=10, # randomly rotate images in the range (degrees, 0 to 180)
zoom_range = 0.1, # Randomly zoom image
width_shift_range=0.1, # randomly shift images horizontally (fraction of total width)
height_shift_range=0.1, # randomly shift images vertically (fraction of total height)
horizontal_flip=False, # randomly flip images
vertical_flip=False) # randomly flip images
datagen.fit(X_train)
history = model.fit_generator(datagen.flow(X_train,Y_train, batch_size=batch_size),
epochs = epochs, validation_data = (X_val,Y_val),
verbose = 2, steps_per_epoch=X_train.shape[0] // batch_size
, callbacks=[learning_rate_reduction])
# 不使用數據增強
history = model.fit(X_train, Y_train, batch_size = batch_size, epochs = epochs,
validation_data = (X_val, Y_val), verbose = 2)
4 評估模型
4.1 訓練和交叉驗證曲線
fig, ax = plt.subplots(2,1)
ax[0].plot(history.history['loss'], color='b', label="Training loss")
ax[0].plot(history.history['val_loss'], color='r', label="validation loss",axes =ax[0])
legend = ax[0].legend(loc='best', shadow=True)
ax[1].plot(history.history['acc'], color='b', label="Training accuracy")
ax[1].plot(history.history['val_acc'], color='r',label="Validation accuracy")
legend = ax[1].legend(loc='best', shadow=True)
- 2個epoch后，accuracy接近99%
- 交叉驗證集准確率基本一直高於訓練集准確率，表明模型沒有過擬合
4.2 混淆矩陣 Confusion matrix
混淆矩陣用於了解模型的缺陷
def plot_confusion_matrix(cm, classes,
normalize=False,
title='Confusion matrix',
cmap=plt.cm.Blues):
"""
This function prints and plots the confusion matrix.
Normalization can be applied by setting `normalize=True`.
"""
plt.imshow(cm, interpolation='nearest', cmap=cmap)
plt.title(title)
plt.colorbar()
tick_marks = np.arange(len(classes))
plt.xticks(tick_marks, classes, rotation=45)
plt.yticks(tick_marks, classes)
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
thresh = cm.max() / 2.
for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
plt.text(j, i, cm[i, j],
horizontalalignment="center",
color="white" if cm[i, j] > thresh else "black")
plt.tight_layout()
plt.ylabel('True label')
plt.xlabel('Predicted label')
# Predict the values from the validation dataset
Y_pred = model.predict(X_val)
# Convert predictions classes to one hot vectors
Y_pred_classes = np.argmax(Y_pred,axis = 1)
# Convert validation observations to one hot vectors
Y_true = np.argmax(Y_val,axis = 1)
# compute the confusion matrix
confusion_mtx = confusion_matrix(Y_true, Y_pred_classes)
# plot the confusion matrix
plot_confusion_matrix(confusion_mtx, classes = range(10))
從圖中可以看出，絕大部分預測是正確的，但仍有個別數字預測錯誤。
接下來調查一下造成錯誤的原因：
# 得到的是bool值，每一個數據是否預測正確
errors = (Y_pred_classes - Y_true != 0)
# 錯誤預測的預測值
Y_pred_classes_errors = Y_pred_classes[errors]
# 錯誤預測上各個數字預測概率
Y_pred_errors = Y_pred[errors]
# 真實值
Y_true_errors = Y_true[errors]
# 對應的圖像
X_val_errors = X_val[errors]
def display_errors(errors_index,img_errors,pred_errors, obs_errors):
""" This function shows 6 images with their predicted and real labels"""
n = 0
nrows = 2
ncols = 3
fig, ax = plt.subplots(nrows,ncols,sharex=True,sharey=True)
for row in range(nrows):
for col in range(ncols):
error = errors_index[n]
ax[row,col].imshow((img_errors[error]).reshape((28,28)))
ax[row,col].set_title("Predicted label :{}\nTrue label :{}".format(pred_errors[error],obs_errors[error]))
n += 1
# Probabilities of the wrong predicted numbers
Y_pred_errors_prob = np.max(Y_pred_errors,axis = 1)
# Predicted probabilities of the true values in the error set
true_prob_errors = np.diagonal(np.take(Y_pred_errors, Y_true_errors, axis=1))
# Difference between the probability of the predicted label and the true label
delta_pred_true_errors = Y_pred_errors_prob - true_prob_errors
# Sorted list of the delta prob errors
sorted_dela_errors = np.argsort(delta_pred_true_errors)
# Top 6 errors
most_important_errors = sorted_dela_errors[-6:]
# Show the top 6 errors
display_errors(most_important_errors, X_val_errors, Y_pred_classes_errors, Y_true_errors)
從圖中可以看出，這些預測錯誤的圖片，確實有很大的迷惑性，對於人來說也不一定能分辨清楚。
5 生成結果
# predict results
results = model.predict(test)
# select the indix with the maximum probability
results = np.argmax(results,axis = 1)
results = pd.Series(results,name="Label")
submission = pd.concat([pd.Series(range(1,28001),name = "ImageId"),results],axis = 1)
submission.to_csv("cnn_mnist_datagen.csv",index=False)