python - 减少 CNN 模型中的验证损失

Question

import tensorflow as tf
import tensorflow.keras
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, Activation, Flatten, Conv2D, MaxPooling2D
import pickle
import numpy as np
from keras.models import model_from_json
from keras.models import load_model
import matplotlib.pyplot as plt

# Opening the files about data
X = pickle.load(open("X.pickle", "rb"))
y = pickle.load(open("y.pickle", "rb"))

# normalizing data (a pixel goes from 0 to 255)
X = X/255.0

# Building the model
model = Sequential()
# 3 convolutional layers
model.add(Conv2D(32, (3, 3), input_shape = X.shape[1:]))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2,2)))

model.add(Conv2D(64, (3, 3)))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2,2)))


model.add(Conv2D(64, (3, 3)))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.9))

# 5 hidden layers
model.add(Flatten())

model.add(Dense(128))
model.add(Activation("relu"))

model.add(Dense(128))
model.add(Activation("relu"))

model.add(Dense(128))
model.add(Activation("relu"))

model.add(Dense(128))
model.add(Activation("relu"))

model.add(Dense(128))
model.add(Activation("relu"))

# The output layer with 7 neurons, for 7 classes
model.add(Dense(13))
model.add(Activation("softmax"))

# Compiling the model using some basic parameters
model.compile(loss="sparse_categorical_crossentropy",
                optimizer="adam",
                metrics=["accuracy"])

# Training the model, with 40 iterations
# validation_split corresponds to the percentage of images used for the validation phase compared to all the images

print("X = " + str(len(X)))
print("y = " + str(len(y)))

history = model.fit(X, y, batch_size=32, epochs=1000, validation_split=0.1)

# Saving the model
model_json = model.to_json()
with open("model.json", "w") as json_file :
    json_file.write(model_json)

model.save_weights("model.h5")

print("Saved model to disk")

model.save('CNN.model')

# Printing a graph showing the accuracy changes during the training phase
print(history.history.keys())

plt.show()

plt.plot(history.history['accuracy'])

plt.plot(history.history['loss'])

plt.title('model accuracy')

plt.ylabel('accuracy')

plt.xlabel('epoch')

plt.legend(['train', 'validation'], loc='upper left')

plt.show()

问题是，我的训练损失较低，但验证准确度很高。而且验证的准确性也极低。我该如何解决这个问题？我试图将下降值增加到 0.9，但损失仍然高得多。我也尝试使用线性函数进行激活，但没有用。

请帮忙。

标签： pythontensorflowmachine-learning

Answer 1

正如已经提到的，在没有看到数据的情况下很难给出好的建议。

我会尝试以下内容： - 在 maxpooling 层之后删除 Dropout - 删除一些密集层 - 在密集层之间添加 dropout

如果仍然过拟合，请在密集层之间添加 dropout

编辑：在我看到损失和准确度图后，我会建议以下内容：

1. 最高优先级是，获取更多数据。
2. 然后使用数据增强甚至增加您的数据集
3. 如果额外的数据没有帮助，进一步降低你的神经网络的复杂性（但我认为训练会随着更多的数据而减慢，并且验证损失也会在更长的时期内减少）