Title:Ensembling Representations for Synergistic Lifelong Learning with Quasilinear Complexity

Abstract:In biological learning, data are used to improve performance not only on the current task, but also on previously encountered, and as yet unencountered tasks. In contrast, classical machine learning starts from a blank slate, or tabula rasa, using data only for the single task at hand. While typical transfer learning algorithms can improve performance on future tasks, their performance on prior tasks degrades upon learning new tasks (called catastrophic forgetting). Many recent approaches for continual or lifelong learning have attempted to maintain performance given new tasks. But striving to avoid forgetting sets the goal unnecessarily low: the goal of lifelong learning, whether biological or artificial, should be to improve performance on both past and future tasks with any new data. Our key insight is that we can ensemble representations learned independently across tasks to transfer omnidirectionally, that is, jointly improve performance on both future unforeseen tasks (forward transfer) and past tasks (backward transfer). Specifically, we propose two lifelong learners: one ensembling trees and the other ensembling networks. In both cases, the algorithms are able to learn synergistically, improving performance on both past and future tasks in a variety of simulated and real data scenarios, including tabular data, image data, spoken data, and adversarial tasks. Moreover, they can do so with quasilinear space and time complexity, a requirement for any bona fide lifelong learning system.