I war reading Class-incremental learning: survey and performance evaluation on image classification. I do not understand clearly what system do we expect at the end. It is 1. or 2.
1.
train a model to predict whether inputs belongs to the class A or B: model 1
train model 1 to predict whether inputs belongs to the class C or D: model 2
or
2.
train a model to predict whether inputs belongs to the class A or B: model 1
train model 1 to predict whether inputs belongs to the class A, B, C or D: model 2
which one is considered when people refers to the class incremental learning.
Please warn me if "data science" is better place to ask this question.
