carrier image

Recurrent Neural Networks for Geometric Problems

Fortunato, Meire,Oriol Vinyals,Navdeep Jaitly

24th International Meshing Roundtable, Elsevier Ltd., pp.Research Note, October 12-14 2015

IMR
PROCEEDINGS

24th International Meshing Roundtable
Austin, TX
October 12-14,2014

University of California, Berkeley, Berkeley, CA, USA
Google Inc., Mountain View, CA, USA
Email: meirefortunato@berkeley.edu, vinyals@google.com, ndjaitly@google.com

Abstract
We introduce a new architecture using Recurrent Neural Networks (RRNs) to learn approximate solutions to three geometric problems - finding planar convex hulls, computing Delaunay triangulations, and the planar Travelling Salesman Problem - using training examples alone. We hope our results on these tasks will encourage a broader exploration of neural learning for discrete geometric problems, including mesh generation.

Download Full Paper (PDF Format)


Contact author(s) or publisher for availability and copyright information on above referenced article