2018 (modified: 11 Nov 2022)ICML 2018Readers: Everyone
Abstract:We introduce a conceptually simple and scalable framework for continual learning domains where tasks are learned sequentially. Our method is constant in the number of parameters and is designed to ...