Publication | Open Access
Progress & Compress: A scalable framework for continual learning
292
Citations
0
References
2018
Year
Artificial IntelligenceIncremental LearningEngineeringMachine LearningSequential LearningEducationRecurrent Neural NetworkContinual Learning DomainsData ScienceRobot LearningContinual Learning (Lifelong Deep Learning)Just-in-time LearningCognitive ScienceSequential ClassificationLearning AnalyticsComputer ScienceDeep LearningNeural Architecture SearchContinual LearningActive ColumnContinual Learning (Educational Psychology)
The study introduces a conceptually simple and scalable framework for continual learning, where tasks are learned sequentially. The framework, called Progress & Compress, keeps a fixed‑size knowledge base and an active column that learns new tasks and then distills them back, preserving prior performance without expanding the network or storing past data.
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 preserve performance on previously encountered tasks while accelerating learning progress on subsequent problems. This is achieved by training a network with two components: A knowledge base, capable of solving previously encountered problems, which is connected to an active column that is employed to efficiently learn the current task. After learning a new task, the active column is distilled into the knowledge base, taking care to protect any previously acquired skills. This cycle of active learning (progression) followed by consolidation (compression) requires no architecture growth, no access to or storing of previous data or tasks, and no task-specific parameters. We demonstrate the progress & compress approach on sequential classification of handwritten alphabets as well as two reinforcement learning domains: Atari games and 3D maze navigation.