Continual Deep Learning
Learning to solve complex sequences of tasks--while transferring knowledge and avoiding catastrophic forgetting--remains a key obstacle to achieving human-level intelligence. Previous methods for continual learning exist, but they are not applicable to deep neural networks. Progressive neural networks are a step forward in this direction: they are immune to forgetting and can leverage prior knowledge via lateral connections to previously learned features. I will explain this new deep learning framework and show how it can be used for transferring knowledge from simulated robotic tasks to a real robot domain.
Raia Hadsell, a senior research scientist at DeepMind, has worked on deep learning and robotics problems for over 10 years. Her thesis on Vision for Mobile Robots won the Best Dissertation award from New York University, and was followed by a post-doc at Carnegie Mellon's Robotics Institute. Raia then worked as a senior scientist and tech manager at SRI International. Raia joined DeepMind in 2014, where she leads a research team studying robot navigation and lifelong learning.