Abstract

Computing with Neural Networks

Alex Graves | Research Scientist | Google DeepMind

Thursday 27 September 2018 | 13:00

Venue | Senior Common Room, Level 2 (2D17), Priory Road Complex

This talk will discuss two related architectures for symbolic computation with neural networks: the Neural Turing Machine and Differentiable Neural Computer.  It will also cover Adaptive Computation Time, a method for dynamically adapting the amount of computation performed by a recurrent neural network to the difficulty of the task. This will hopefully lead to a fruitful discussion on the future of neural networks for learning programs and embodying algorithmic forms of generalisation.

All Welcome | Tea, coffee and biscuits will be available after the seminar.