mk
Programming with a Differentiable Forth Interpreter
Matko Bosnjak (1), Tim Rocktaschel (2), Jason Naradowsky (3) und Sebastian Riedel (1) haben im Juli 2017 im arXiv ihre Forschungsergebnisse veröffentlicht als Proceedings of the 34 th International Conference on Machine Learning, Sydney, Australia. Sie fande sich in England zusammen: (1) Department of Computer Science, University College London, London, UK; (2) Department of Computer Science, University of Oxford, Oxford, UK; (3) Department of Theoretical and Applied Linguistics, University of Cambridge, Cambridge, UK. Forth also nicht nur in Wien!
"In this approach, the programmer specifies a program sketch ... in a traditional programming language. This sketch defines one part of the neural network behaviour. The other part is learned using training data. The core insight that enables this approach is the fact that most programming languages can be formulated in terms of an abstract machine that executes the commands of the language. We implement these machines as neural networks, constraining parts of the networks to follow the sketched behaviour. ... "