带有示例图像的CIFAR-10类
深度学习隐喻:将ConvNet层比作Jenga块
model = Sequential()
model.add(Flatten(input_shape=(32, 32, 3)))
model.add(Dense(10))
model.add(Activation('softmax'))
model.compile(optimizer=’adam’,
loss='categorical_crossentropy',
metrics=['accuracy'])
OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)
Input ##### 32 32 3
Flatten ||||| ------------------- 0 0.0%
##### 3072
Dense XXXXX ------------------- 30730 100.0%
softmax ##### 10
$ neptune send lr.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium
Neptune通道仪表盘中显示的错误分类的图像
model = Sequential()
model.add(Flatten(input_shape=(32, 32, 3)))
model.add(Dense(128, activation='sigmoid'))
model.add(Dense(128, activation='sigmoid'))
model.add(Dense(10))
model.add(Activation('softmax'))
model.compile(optimizer=adam,
loss='categorical_crossentropy',
metrics=['accuracy'])
OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)
Input ##### 32 32 3
Flatten ||||| ------------------- 0 0.0%
##### 3072
Dense XXXXX ------------------- 393344 95.7%
sigmoid ##### 128
Dense XXXXX ------------------- 16512 4.0%
sigmoid ##### 128
Dense XXXXX ------------------- 1290 0.3%
softmax ##### 10
$ neptune send mlp.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium
训练集和验证集的准确性和log-loss
model = Sequential()
model.add(Conv2D(32, (3, 3), activation='relu',
input_shape=(32, 32, 3)))
model.add(MaxPool2D())
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPool2D())
model.add(Flatten())
model.add(Dense(10))
model.add(Activation('softmax'))
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)
Input ##### 32 32 3
Conv2D \|/ ------------------- 896 2.1%
relu ##### 30 30 32
MaxPooling2D Y max ------------------- 0 0.0%
##### 15 15 32
Conv2D \|/ ------------------- 18496 43.6%
relu ##### 13 13 64
MaxPooling2D Y max ------------------- 0 0.0%
##### 6 6 64
Flatten ||||| ------------------- 0 0.0%
##### 2304
Dense XXXXX ------------------- 23050 54.3%
softmax ##### 10
$ neptune send cnn_simple.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium
OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)
Input ##### 32 32 3
Conv2D \|/ ------------------- 896 0.1%
relu ##### 32 32 32
Conv2D \|/ ------------------- 1056 0.2%
relu ##### 32 32 32
MaxPooling2D Y max ------------------- 0 0.0%
##### 16 16 32
BatchNormalization μ|σ ------------------- 128 0.0%
##### 16 16 32
Dropout | || ------------------- 0 0.0%
##### 16 16 32
Conv2D \|/ ------------------- 18496 2.9%
relu ##### 16 16 64
Conv2D \|/ ------------------- 4160 0.6%
relu ##### 16 16 64
MaxPooling2D Y max ------------------- 0 0.0%
##### 8 8 64
BatchNormalization μ|σ ------------------- 256 0.0%
##### 8 8 64
Dropout | || ------------------- 0 0.0%
##### 8 8 64
Conv2D \|/ ------------------- 73856 11.5%
relu ##### 8 8 128
Conv2D \|/ ------------------- 16512 2.6%
relu ##### 8 8 128
MaxPooling2D Y max ------------------- 0 0.0%
##### 4 4 128
BatchNormalization μ|σ ------------------- 512 0.1%
##### 4 4 128
Dropout | || ------------------- 0 0.0%
##### 4 4 128
Flatten ||||| ------------------- 0 0.0%
##### 2048
Dense XXXXX ------------------- 524544 81.6%
relu ##### 256
Dropout | || ------------------- 0 0.0%
##### 256
Dense XXXXX ------------------- 2570 0.4%
softmax ##### 10
$ neptune send cnn_adv.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium