Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
Imagine you want to predict if a picture is showing a cat or not. First you train your ML algorithm with examples of pictures of cats and dogs and it works. But then you want to train it to also differentiate between cats and lions, and cats and pumas (for example). My question is: would the distribution of examples be 50% cats and 50% everything else (33% dogs, 33% lions and 33% pumas) or 25% cats, 25% dogs, 25% lions and 25% pumas?
Sorry if my question is too basic. Thanks for the answers!
I would worry less about the split and more about getting a good representation of pictures of the animals. Otherwise the model might end up predicting snow when you think it is predicting wolves.
In addition, many times you do not get that chance to balance how you want to. The real-world comes into play. Is your model predicting animals in that balance? Just because a training set is balanced one way, does not necessarily mean the model is good or bad. For some views, take a look at the comments.
