Convolution layer dimension mismatch issue-CodePudding

I'm trying to learn how the convolution layer works in neural networks. I found two different related posts for a similar issue and tried the suggestions, but could not get around.

import tensorflow as tf
model_valid = tf.keras.Sequential([
    tf.keras.layers.Flatten(input_shape=(10,)),
    tf.keras.layers.Dense(16,  activation='relu'),
    tf.keras.layers.Conv1D(16, kernel_size=(2), activation='relu', padding='same'),     
    tf.keras.layers.MaxPooling1D(pool_size=(4), strides=3, padding='valid'),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(1, activation='softmax')
])
model_valid.summary()

I am receiving an incompatiblity issue as Input 0 of layer conv1d_37 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [None, 16]. The issue arises when building up the convolution layer.

CodePudding user response：

The input's shape before the first Conv1D is incorrect. As user spb suggested, the input to Conv1D must be a 3D tensor, though removing the flatten layer (as suggested) won't fix the problem.

Instead, simply add an extra dimension after the first Dense layer.

import tensorflow as tf
model_valid = tf.keras.Sequential([
    tf.keras.layers.Flatten(input_shape=(10,)),
    tf.keras.layers.Dense(16,  activation='relu'),
    tf.keras.layers.Reshape((16, 1)), # ADD THIS LINE OF CODE
    tf.keras.layers.Conv1D(16, kernel_size=(2), activation='relu', padding='same'),     
    tf.keras.layers.MaxPooling1D(pool_size=(4), strides=3, padding='valid'),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(1, activation='softmax')
])
model_valid.summary()

Now the model summary is:

Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
flatten_4 (Flatten)          (None, 10)                0
_________________________________________________________________
dense_4 (Dense)              (None, 16)                176
_________________________________________________________________
reshape (Reshape)            (None, 16, 1)             0
_________________________________________________________________
conv1d_1 (Conv1D)            (None, 16, 16)            48
_________________________________________________________________
max_pooling1d (MaxPooling1D) (None, 5, 16)             0
_________________________________________________________________
flatten_5 (Flatten)          (None, 80)                0
_________________________________________________________________
dense_5 (Dense)              (None, 1)                 81
=================================================================
Total params: 305
Trainable params: 305
Non-trainable params: 0
_________________________________________________________________

CodePudding user response：

Input to Conv1D layer must be 3D Tensor (batch_size, steps, input_dim). Flattening the tensor will convert input into 2D tensor (batch_size, input_dim). Try removing the Flatten layer.