Can i replace all dense layer with convolutional layer in neural networks and still get same output

Course Q&A Machine Learning Specialization Advanced Learning Algorithms
1

I was wondering whether we could replace all dense layer applied approach to a solution with a convolutional layer and still get the same output or not. So when exactly we should use dense layer and convolutional layer or any other layer.

2

Probably not the same results. They perform different transformations on data.

Convolutions are generally good for images and dense layers for other kind of data, like tabular data lets say…

There are other usages and other types of layers too.

1 Like
3

Hello @Harinandan,

It depends on your application. You could get a semi-similar result. But as i know, convolutional layers are mostly good for when you want to find a pattern in the input.

Cheers,
Amir