Each example is a random vector in the same “space” as the sample in the top left.
Each row represents one dimension of our vectors. You can drag them back and forth to change the value of our vector for that dimension.
The bar below each vector is a measure of how similar it is to the sample in the top left.
Cosine similarity is a technique for getting the “distance” between two vectors in high-dimensional spaces. We intuitively understand distance in 2 and 3 dimensions, but what happens if we have more than that? It can be convenient to calculate a single number that tells us how similar or different two high-dimensional vectors are.
I made this explanation because I’m trying to use find similar blocks and in order to do that I need to understand cosine similarity. They say nothing helps you understand better than trying to explain it!