Keep each picture together

A journey through the shape guarantee with Botie.

Neural Networks · A journey with Botie

First visit? Learn the workshop controls

Botie's notebook

How Does a Grid of Features Reach a Final Decision?

Part 1 of 3

An image-processing layer may produce several grids, one for each detected feature. A later layer may expect a single list of values per example. These are different organisations of numbers, so connecting them requires an explicit agreement about dimensions and ordering.

Sources for this history

LeCun and colleagues (1998), Gradient-Based Learning Applied to Document Recognition

← Previous lesson: A Dense Layer and the Forward Pass Read the full lesson and derivations