When two neural networks reach nearly identical accuracy, have they actually learned the same thing? The Problem With Calling Two Models "Equivalent" Imagine training two neural networks on the same classification problem. Both eventually achiev...