Task 001/002: Neuro-Inspired Algorithms for Efficient and Lifelong Learning/Theoretical Underpinnings of Neuro-inspired Computing

Abstract:"The human brain is the gold standard of adaptive learning. It not only can learn and benefit from experience, but also can adapt its behavior to new situations. In contrast, deep neural networks only learn one sophisticated but fixed mapping from inputs to outputs. This limits their applicability to more complex and dynamic situations, in which the input to output mapping may change with different tasks or contexts. Here, inspired by the human brain, we propose a new kind of deep neural network with extra, out-of-network, task-dependent biasing units that allow, for the first time, a single network to learn multiple parallel input to output mappings and to switch on the fly between them at runtime. Biasing units are programmed by leveraging beneficial perturbations (opposite to well-known adversarial perturbations) tailored to each task. One way to demonstrate efficacy of the approach is on the problem of continual learning. Continual learning of multiple tasks in artificial neural networks using gradient descent leads to catastrophic forgetting, whereby a previously learned mapping of an old task is erased during learning of a new mapping for a new task. In the proposed approach, beneficial perturbations for a given task bias the network toward that task, essentially switching the network into a different mode to process that task. This largely eliminates catastrophic interference between tasks. Our approach is more parameter-efficient (additional 0.3\% parameters per task) and more memory-efficient (does not need to store any data from previous tasks) compared with previous continual learning paradigms."

Bio: "Shixian Wen is a 3-year Ph.D. student at the University of Southern California. He is interested in combining neuroscience with computer science to develop advanced bio-plausible machine learning  algorithms and better understand of human brain."