DynMat, a network that can learn after learning
This addresses the challenge of continuous, online learning for AI agents, offering a potential step towards more brain-like intelligence, though it appears incremental as it builds on existing CLS theory.
The authors tackled the problem of catastrophic interference in artificial neural networks by proposing DynMat, a dual-system architecture inspired by the brain's complementary learning systems, which enables learning new classes without forgetting and reduces the need for offline training.
To survive in the dynamically-evolving world, we accumulate knowledge and improve our skills based on experience. In the process, gaining new knowledge does not disrupt our vigilance to external stimuli. In other words, our learning process is 'accumulative' and 'online' without interruption. However, despite the recent success, artificial neural networks (ANNs) must be trained offline, and they suffer catastrophic interference between old and new learning, indicating that ANNs' conventional learning algorithms may not be suitable for building intelligent agents comparable to our brain. In this study, we propose a novel neural network architecture (DynMat) consisting of dual learning systems, inspired by the complementary learning system (CLS) theory suggesting that the brain relies on short- and long-term learning systems to learn continuously. Our experiments show that 1) DynMat can learn a new class without catastrophic interference and 2) it does not strictly require offline training.