Yes Jane, you are right. Suppose we have k(no of neighbors hyper parameter) = 5.

• As you said first for k = 1, train on 90% data and predicts rest 10%. then repeats that 9 more times each with a different 10% of data heldout, then the average of those 10 prediction trial is stored for k =1
• This is repeated from k=1 to k= 5