Learning and analyzing vector encoding of symbolic representations

03/10/2018

∙

by   Roland Fernandez, et al.

∙

Microsoft

∙

Johns Hopkins University

∙

0

∙

share

We present a formal language with expressions denoting general symbol structures and queries which access information in those structures. A sequence-to-sequence network processing this language learns to encode symbol structures and query them. The learned representation (approximately) shares a simple linearity property with theoretical techniques for performing this task.