Hi Maher,

Thanks for the article! I was wondering if there is any industrial accepted good enough precision at k threshold? I understand it really depends on the context and if it’s cancer diagnoses then the precision better be super high. I am mostly interested in the ecommerce website recommending products to customer setting, where a not so relevant recommendation is not devastating. I am asking because we have a lot of customers and products and searching in the hyperparameter space is taking very long and I am not sure how much effort we should be putting into the training before we call it “good enough”.

Thanks!

Tina