Cross Validation Hold-out & K-Fold

[dev-fredericfox]

The documentation doesn't really state what type of Cross Validation brain is doing. But since there is an optional K value, I assume it's a K-Fold. What is the default value if K is omitted? Or is it determined based on the dataset?

Is there a way to train on 80% of the data and cross validate with remaining 20% (sometimes called Hold-out)?

[robertleeplummerjr]

The documentation doesn't really state what type of Cross Validation brain is doing. But since there is an optional K value

That is correct, it is K-Fold

What is the default value if K is omitted?:

4 as stated here -

brain.js/src/cross-validate.ts

Line 184 in 5d523fe

k = 4

Is there a way to train on 80% of the data and cross validate with remaining 20% (sometimes called Hold-out)?

I don't believe so. The data is essentially split into k folds, and trained on, and then later compared.