1
$\begingroup$

It is very common for neural networks to be asymmetric about the x axis, that is, to have many more nuerons in the first few layers than in the last few layers. Common example:

enter image description here

But can neural networks be structured in such a way that they're asymmetric about the y axis, that is, such that some parts of the network are 'deeper' than others?

An example of where this could be useful is where a problem involves some unsophisticated features (that require only a small number of layers), and some much more sophisticated features (which may benefit greatly from having more layers - i.e. greater 'depth').

$\endgroup$
1
$\begingroup$

"Wide and Deep network" was used in a 2016 paper.



enter image description here
$\hspace{5cm}$Wide & Deep Learning for Recommender Systems


You may create using keras Concatenate layer

concat = keras.layers.Concatenate()([input, hidden2])

References to read -
The paper
Google blog
Another paper

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.