Filters focus on connectivity between adjacent pixels, because of their fixed size. This means that they will most strongly learn local features. When combined with other filters in layers, and nonlinearities, this allows us to gradually pay attention to larger and more complex features. This stacking of localized features is really desirable and a key reason why convolutional layers are so great.
Local connectivity
Get Deep Learning Quick Reference now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
