Recurrent Neural Networks and Models of Computation - Edward Grefenstette, DeepMind

This talk presents an analysis of various recurrent neural network architectures in terms of traditional models of computation. It makes the case for simpler recurrent architectures being closer to finite state automata, and argues that memory-enhanced architectures support better algorithmic efficiency, even in problems which are describable as regular languages. Logic has proved in the last decades a powerful tool in understanding complex systems. It is instrumental in the development of formal methods, which are mathematically based techniques obsessing on hard guarantees. Learning is a pervasive paradigm which has seen tremendous success recently. The use of statistical approaches yields practical solutions to problems which yesterday seemed out of reach. These two mindsets should not be kept apart, and many efforts have been made recently to combine the formal reasoning offered by logic and the power of learning. The goal of this workshop is to bring together expertise from various areas to try and understand the opportunities offered by combining logic and learning. The programme has four axes, starting from a theoretical standpoint and going to a more practical one: logic and automata, verification, programming languages, and neural networks. Registration for this event is now closed, but if you would like to be added to the waiting list please email: events@turing.ac.uk