Flattening and the Full Connection Operation
In this lesson, we are going to learn about the two remaining steps in implementing the Convolutional Neural Network: flattening and full connection operation.
We'll cover the following...
What is “flattening”?
This is a very simple step. After finishing the previous steps, we obtain a pooled feature map. As the name of this step implies, we are basically going to flatten the pooled feature map into a column vector as shown below.
Explanation:
- Take a look at the pooled feature map. This was the matrix that we obtained after applying the
max poolingoperation. - We are essentially converting this matrix to a single column vector. We are simply taking one row