Limit on number of categories in classification models

In classification models, is there a limit on the number of categories? Does “discrete and finite” imply that when classifying incoming documents using Supplier names and a company works with 2000 suppliers, the model would fail?

It would not fail, there are classification models with 1000 classes like imagenet. Of course the training of the model might take more time and model itself might become more complex. It will also require more training data.

What could also be done is the algorithm used for face recognition. In theory also a classification, but technically done as some kind of similarity problem. There is also a course from Andrew Ng where this is mentioned, not sure of the name of the course, maybe someone can help out

Hi @Markus_Degen, I think you are talking about the Siamese network, which is covered in Professor Ng’s Deep Learning Specialization Course 4 Week 4.


Exactly that @rmwkwok . Perfect :slight_smile:

With this you are not limited to the number of classes