ValueError：无法将大小为 78543360 的数组重塑为形状 (51135,4,32,32)

Question

遵循来自https://machinelearningmastery.com/how-to-develop-rnn-models-for-human-activity-recognition-time-series-classification/的本教程

我正在尝试在我的时间序列数据上实现 CNN-LSTM 网络模型。

verbose, epochs, batch_size = 0, 25, 64
n_timesteps, n_features, n_outputs = 
X_train.shape[1], X_train.shape[2], y_train.shape[1]
# reshape data into time steps of sub-sequences
n_steps, n_length = 4, 32
X_train = X_train.reshape((X_train.shape[0], n_steps, 
n_length, n_features))
X_test = X_test.reshape((X_test.shape[0], n_steps, 
n_length, n_features))
# define model
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, 
kernel_size=3, activation='relu'), input_shape= 
(None,n_length,n_features)))
model.add(TimeDistributed(Conv1D(filters=64, 
kernel_size=3, activation='relu')))
model.add(TimeDistributed(Dropout(0.5)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dropout(0.5))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='sigmoid'))
model.compile(loss='binary_crossentropy', 
optimizer='adam', metrics=['accuracy'])
model.fit(X_train, y_train, epochs=epochs, 
batch_size=batch_size, verbose=verbose)

以下是使用 reshape 之前的形状：

print(X_train.shape, y_train.shape, X_test.shape, 
y_test.shape)
(51135, 128, 12) (51135, 1) (21915, 128, 12) (21915, 1)

这是我遇到错误的代码部分： ValueError: cannot reshape array of size 78543360 into shape (51135, 4,32,32)。

我该如何解决这个问题？

Answer 1

我无权访问您的数据，因此我为具有您指定的维度的数据创建了虚拟 np.zero 数组。运行它以查看维度发生了什么。没有遇到你描述的错误。我使用的代码如下。跑步和训练，使尺寸看起来正确。

X_train=np.zeros((51135, 128, 12))
y_train=np.zeros((51135, 1))
X_test=np.zeros((21915, 128, 12))
y_test=np.zeros((21915, 1))
print(X_train.shape, y_train.shape, X_test.shape,y_test.shape)
verbose, epochs, batch_size = 0, 25, 64
n_timesteps, n_features, n_outputs = X_train.shape[1], X_train.shape[2], y_train.shape[1]
print(n_timesteps, n_features, n_outputs)
# reshape data into time steps of sub-sequences
n_steps, n_length = 4, 32
X_train = X_train.reshape((X_train.shape[0], n_steps, n_length, n_features))
X_test = X_test.reshape((X_test.shape[0], n_steps,n_length, n_features))
print (X_train.shape, X_test.shape)
# define model
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, 
kernel_size=3, activation='relu'), input_shape= 
(None,n_length,n_features)))
model.add(TimeDistributed(Conv1D(filters=64, 
kernel_size=3, activation='relu')))
model.add(TimeDistributed(Dropout(0.5)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dropout(0.5))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='sigmoid'))
model.compile(loss='binary_crossentropy',optimizer='adam', metrics=['accuracy'])
model.summary()
model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, verbose=1)

这是打印出来的数据。当然，全零训练数据是没有意义的，但维度适合 model.fit 执行

(51135, 128, 12) (51135, 1) (21915, 128, 12) (21915, 1)
128 12 1
(51135, 4, 32, 12) (21915, 4, 32, 12)
Epoch 1/25
799/799 [==============================] - 9s 11ms/step - loss: 0.6051 - accuracy: 1.0000
Epoch 2/25
799/799 [==============================] - 5s 6ms/step - loss: 0.3479 - accuracy: 1.0000

Answer 2

您正在尝试将 X_train 重塑为 (51135, 4,32,32)。这是不可能的，因为 X_train 的形状是 (51135, 128, 12)。 51135 x 128 x 12 ！= 51135 x 4 x 32 x 32