RNNs can generate bounded hierarchical languages with optimal memory.

John Hewitt Michael Hahn Surya Ganguli Percy Liang Christopher D. Manning

Published in: EMNLP (1) (2020)

Keyphrases

recurrent neural networks
expressive power
optimal solution
dynamic programming
bounded memory
worst case
memory size
database
language independent
computing power
feedforward neural networks
asymptotically optimal
limited memory
automatically generate
automatically generating
cross lingual
coarse to fine
data sets