ValueError：预期 min_ndim=3，发现 ndim=2

Question

我正在尝试将输入分类。

形状是：

df_train.shape: (17980, 380)
df_validation.shape: (17980, 380)

但是，当我运行我的代码时，我收到以下错误

ValueError: Input 0 of layer conv1d is incompatible with the layer: : expected min_ndim=3, found ndim=2. Full shape received: [32, 380]

我们如何解决这个错误？

Answer 1

Conv1D 接受形状的输入：

3+D 张量，形状为：batch_shape + (steps, input_dim)

如果您的数据只是 2D，请添加一个虚拟维度：

df_train = df_train[..., None]
df_validation = df_validation[..., None]

还相应地修改 batch_input_shape=(32, 1, 380) 为：batch_input_shape=(32, 380, 1) 或者完全省略它

其他更改（处理此虚拟数据）：

df_train = np.random.normal(size=(17980, 380))
df_validation = np.random.normal(size=(17980, 380))

df_train = df_train[..., None]
df_validation = df_validation[..., None]

y_train = np.random.normal(size=(17980, 1))
y_validation = np.random.normal(size=(17980, 1))

#train,test = train_test_split(df, test_size=0.20, random_state=0)


    
batch_size=32
epochs=5
    
model = Sequential()

model.add((Conv1D(filters=5, kernel_size=2, activation='relu', padding='same')))
model.add((MaxPooling1D(pool_size=2)))
model.add(LSTM(50, return_sequences=True))
model.add(LSTM(10))
model.add(Dense(1))

adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0)

model.compile(optimizer=adam, loss='mse', metrics=['mae', 'mape', 'acc'])
callbacks = [EarlyStopping('val_loss', patience=3)]

model.fit(df_train, df_validation, batch_size=batch_size)

print(model.summary())