I'm trying to implement KNN classification with cross-validation implementation in python. The data consists of 10 folds of size 99x64, each with their corresponding label of size 99x1. Do I have to calculate distances row by row between each fold, for a resulting $99x1$ distance-vector for every $k-1$ testing fold? E.g. Let's say we have three folds, each of dimension $[99x64]$. Assuming we are testing on fold $1$ with folds 2, 3 being the validating datasets, I would end up with two distance-vectors of size $[99x1]$ since I'm calculating the distance between rows of each fold.