1 paper
Dongkyu Cho, Taesup Moon, Rumi Chunara +2
Continual learning (CL) has traditionally focused on minimizing exemplar memory, a constraint often misaligned with modern systems where GPU time, not storage, is the primary bottl…