Some Confusion About n_one In FHDDM

[9sea]

The variable n_one is in river.drift.binary.fhddm.
My personal understanding is that a correct prediction is typically marked as 1, while an incorrect one is marked as 0, which is also mentioned in the original paper. However, in the code, it seems that a correct prediction is marked as 0 and an incorrect prediction as 1, which is the opposite of what I expected. This has caused me some confusion, and I would appreciate it if someone could clarify this for me. Thank you!

[smastelini]

Hi @9sea, following the same patterns as those used in the reference MOA implementation, the inputs refer to a 0-1 loss. So, 0 means no error and 1 represents a misclassification.

[9sea]

Hi, @smastelini , I did not find fhddm in MOA, but I find another realization as picture shown.

[smastelini]

My bad, I missed the "F" at the beginning of the method name. I am pinging @gabrieljaguiar who did the implementation.

[9sea]

I modify "update" function, Insert the line “x = 0 if x else 1” at the first line of the function， then I do some easy experiments, it has better performance.
Thank you for your answer!

[gabrieljaguiar]

@9sea n_one is implemented as the number of correct predictions.

[9sea]

@gabrieljaguiar The parameter x of the function update is 1, indicating a prediction error, and 0 indicating a correct prediction.