Which lesson had a snippet on nearest neighbor image


I’m sure I remember in one of the videos Jeremy did a quick demo of classifying an image of an unknown class and finding its nearest neighbour from the trained classes. I can’t find it in the part 1 or 2 timelines. Can anyone point me at it??

I figure it was training image classification with embeddings rather than one-hot encoding. Just can’t remember the details :frowning:



In Lesson 11—multi-modal learning, when Jeremy started covering the DeViSE paper (see my lesson notes for this)

Or in particular, look for the “Nearest neighbor search” key concept under the same DeViSE topic in my lesson notes (or the exact location in lesson 11 video timeline).

Hope it helps.