What are the practical example of different Continual Learning Scenarios?

I have been going through a few of the survey papers, and each survey papers define the meaning of Domain incremental learning, class incremental learning, and task incremental learning differently. I am kind of confused between these 3 topics.

Joining the group lateā€¦

Domain incremental learning: The incremental data contains data shifted versions of the source dataset (used to train the baseline source model). E.g. of this would be covariate shifted data. We can consider this an incrementally applied domain adaptation setting with minimal catastrophic forgetting. The feature space and label set is assumed to be the same.

Class incremental learning: The incremental target data contains target private classes apart from all the shared source classes, i.e., C_target = C_shared U C_private (open-set problem setting). In every iteration, we add new classes to the baseline model. The feature space can be assumed to be the same.

In my opinion, class incremental learning can be considered to be synonymous or a subset of task incremental learning. Class incremental learning is mostly used in a domain adaptation setting, whereas, task incremental learning is more generic. In a task incremental learning setting, the source and target label sets can be non-overlapping.