Exemplar-free Continual Representation Learning via Learnable Drift Compensation | Read Paper on Bytez