From the course: Neural Networks and Convolutional Neural Networks Essential Training
Join today to access over 25,300 courses taught by industry experts.
Zero padding and pooling
From the course: Neural Networks and Convolutional Neural Networks Essential Training
Zero padding and pooling
Our detective team has now examined the entire data with their magnifying glasses, producing detailed feature maps. But here's the problem, we have too much evidence. Every tiny detail is documented, creating massive amounts of data to process. So what if we could summarize the evidence without losing the important findings? This is where Max Spooling can help. Think of this as a senior detective reviewing all of the evidence and keeping only the strongest findings from each area. So here's how it works. So you divide each feature map into two-by-two blocks, and from each block of our four values, keep only the maximum, the strongest detection. Now, the maximum value in the top left corner is seven. So slide to the next two-by-two block with a stride of two, so there's no overlap. Take the maximum again, and this time it's four. Now move to the bottom left, and so on. And the result is your 4x4 feature map becomes 2x2. You've reduced it by half in each dimension, keeping only the most…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.