• Publications
  • Influence
Overcoming catastrophic forgetting in neural networks
It is shown that it is possible to overcome the limitation of connectionist models and train networks that can maintain expertise on tasks that they have not experienced for a long time and selectively slowing down learning on the weights important for previous tasks. Expand
Progressive Neural Networks
This work evaluates this progressive networks architecture extensively on a wide variety of reinforcement learning tasks, and demonstrates that transfer occurs at both low-level sensory and high-level control layers of the learned policy. Expand
Neural scene representation and rendering
The Generative Query Network (GQN) is introduced, a framework within which machines learn to represent scenes using only their own sensors, demonstrating representation learning without human labels or domain knowledge. Expand
Vector-based navigation using grid-like representations in artificial agents
These findings show that emergent grid-like representations furnish agents with a Euclidean spatial metric and associated vector operations, providing a foundation for proficient navigation, and support neuroscientific theories that see grid cells as critical for vector-based navigation. Expand
On the importance of single directions for generalization
It is found that class selectivity is a poor predictor of task importance, suggesting not only that networks which generalize well minimize their dependence on individual units by reducing their selectivity, but also that individually selective units may not be necessary for strong network performance. Expand
Machine Theory of Mind
It is argued that this system -- which autonomously learns how to model other agents in its world -- is an important step forward for developing multi-agent AI systems, for building intermediating technology for machine-human interaction, and for advancing the progress on interpretable AI. Expand
Human-level performance in 3D multiplayer games with population-based reinforcement learning
A tournament-style evaluation is used to demonstrate that an agent can achieve human-level performance in a three-dimensional multiplayer first-person video game, Quake III Arena in Capture the Flag mode, using only pixels and game points scored as input. Expand
Attention stabilizes the shared gain of V4 populations
A functional model of population activity of bilateral neural populations in area V4 reveals four separate time-varying shared modulatory signals, the dominant two of which each target task-relevant neurons in one hemisphere are dominant. Expand
The Predictron: End-To-End Learning and Planning
The predictron consists of a fully abstract model, represented by a Markov reward process, that can be rolled forward multiple "imagined" planning steps that accumulates internal rewards and values over multiple planning depths. Expand
Constructing Noise-Invariant Representations of Sound in the Auditory Pathway
Along the auditory pathway from auditory nerve to midbrain to cortex, individual neurons adapt progressively to sound statistics, enabling the discernment of foreground sounds, such as speech, overExpand