Normative Loss Landscape Navigation: A Trajectory-Based Approach to Mitigating Forgetting in Incremental Learning
TMLN (Trajectory-Modulatory Landscape Navigation), a normative navigation policy that formalizes continual learning as an optimal control problem over a curved loss landscape, is proposed and significantly reduces the loss barrier between consecutive tasks.
I. Aguilar, Z. A. Zainal, Luis Fernando Herbozo Contreras et al.
· 0 citations