I am looking to train a binary classifier. Most of my experience so far has been with generative models, not classifiers, so I am wondering with respect to training data, what is a good ratio of 0 and 1 classes? In my case there are significantly more points I want the network to reject (0) than accept (1). Is 10:1 an unreasonable ratio? 100:1? (I have a fixed amount of points in the 1-class (~180k datapoints) which would be difficult to augment, but I am able to easily generate as many 0-class points as I need).
Asked
Active
Viewed 47 times
