参考链接:https://blog.goodaudience.com/introduction-to-1d-convolutional-neural-networks-in-keras-for-time-sequences-3a7ff801a2cf
何时使用 1D-CNN
- 从短(固定长度)片段内提取特征
- 片段内特征位置没有相关性
A 1D CNN is very effective when you expect to derive interesting
features from shorter (fixed-length) segments of the overall data set
and where the location of the feature within the segment is not of
high relevance.
适用数据: 传感器时序数据
1D-CNN 与 2D-CNN 的区别
- 输入数据的维度不同
- 卷积遍历数据的方式不同
应用:行为识别
- 加速计数据:x, y, z 三轴
- 数据类别:走、慢跑、站立等
构造 1D-CNN
Keras 构造网络:
model_m = Sequential()
model_m.add(Reshape((TIME_PERIODS, num_sensors), input_shape=(input_shape,)))
model_m.add(Conv1D(100, 10, activation='relu', input_shape=(TIME_PERIODS, num_sensors)))
model_m.add(Conv1D(100, 10, activation='relu'))
model_m.add(MaxPooling1D(3))
model_m.add(Conv1D(160, 10, activation='relu'))
model_m.add(Conv1D(160, 10, activation='relu'))
model_m.add(GlobalAveragePooling1D())
model_m.add(Dropout(0.5))
model_m.add(Dense(num_classes, activation='softmax'))
print(model_m.summary())
网络结构
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
reshape_45 (Reshape) (None, 80, 3) 0
_________________________________________________________________
conv1d_145 (Conv1D) (None, 71, 100) 3100
_________________________________________________________________
conv1d_146 (Conv1D) (None, 62, 100) 100100
_________________________________________________________________
max_pooling1d_39 (MaxPooling (None, 20, 100) 0
_________________________________________________________________
conv1d_147 (Conv1D) (None, 11, 160) 160160
_________________________________________________________________
conv1d_148 (Conv1D) (None, 2, 160) 256160
_________________________________________________________________
global_average_pooling1d_29 (None, 160) 0
_________________________________________________________________
dropout_29 (Dropout) (None, 160) 0
_________________________________________________________________
dense_29 (Dense) (None, 6) 966
=================================================================
Total params: 520,486
Trainable params: 520,486
Non-trainable params: 0
_________________________________________________________________
None
- 输入数据:
每个样本为3轴数据,每轴包含4秒数据,采样频率为20Hz,即每轴有80个数值,共240个数值 - Reshape层: 将一个样本的240个数值转为 80×3 矩阵
- 第一个Conv1D层:
- 卷积核长度为10,深度(通道数)为3,步长为1,卷完后数据由 80×3 变为 71×1
- 共100个卷积核,输出 71×100
- 参数个数:每个卷积核 10×3+1=31 个参数,100个卷积核共 3100 个参数
- 第二个Conv1D层:
- 卷积核长度为10,深度(通道数)为100,步长为1,卷完后数据由 71×100 变为 62×1
- 共100个卷积核,输出 62×100
- 参数个数:每个卷积核 10×100+1=101 个参数,100个卷积核共 10100 个参数
- MaxPooling层:
-
池化核长度为3,步长为3, 池化后数据长度为 (62-3) / 3 + 1 = 20
[1,2,3] [4,5,6] [ ... ] [58,59,60] [61,62 ↓ ↓ ↓ max1 max2 ... max20
数据深度(通道数)为100,输出 20×100
-
- 第三个Conv1D层:
- 卷积核长度为10,深度为100,步长为1,卷积核个数为160,输出为
(20-10+1)×160=11×160 - 参数个数:(10×100+1)×160=160160
- 卷积核长度为10,深度为100,步长为1,卷积核个数为160,输出为
- 第四个Conv1D层:
- 卷积核长度为10,深度为160,步长为1,卷积核个数为160,输出为
(11-10+1)×160=2×160 - 参数个数:(10×160+1)×160=256160
- 卷积核长度为10,深度为160,步长为1,卷积核个数为160,输出为
- GlobalAveragePooling1D层:
- 每个通道取均值,共160个通道,输出 160 个数值
- Dropout层
- Dense层(输出层):
- 输入为160个数值,输出为6个数值,激活函数为 softmax
- 参数个数:(160+1)×6=966
训练
callbacks_list = [
keras.callbacks.ModelCheckpoint(
filepath='best_model.{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss', save_best_only=True),
keras.callbacks.EarlyStopping(monitor='acc', patience=1)
]
model_m.compile(loss='categorical_crossentropy',
optimizer='adam', metrics=['accuracy'])
BATCH_SIZE = 400
EPOCHS = 50
history = model_m.fit(x_train,
y_train,
batch_size=BATCH_SIZE,
epochs=EPOCHS,
callbacks=callbacks_list,
validation_split=0.2,
verbose=1)