pith. sign in

arxiv: 1711.05408 · v2 · pith:4AICTXRDnew · submitted 2017-11-15 · 💻 cs.FL · cs.CC· cs.CL

Recurrent Neural Networks as Weighted Language Recognizers

classification 💻 cs.FL cs.CCcs.CL
keywords rnnsapplicationsbecomeslanguagelengthnetworksneuralproblem
0
0 comments X p. Extension
pith:4AICTXRD Add to your LaTeX paper What is a Pith Number?
\usepackage{pith}
\pithnumber{4AICTXRD}

Prints a linked pith:4AICTXRD badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

read the original abstract

We investigate the computational complexity of various problems for simple recurrent neural networks (RNNs) as formal models for recognizing weighted languages. We focus on the single-layer, ReLU-activation, rational-weight RNNs with softmax, which are commonly used in natural language processing applications. We show that most problems for such RNNs are undecidable, including consistency, equivalence, minimization, and the determination of the highest-weighted string. However, for consistent RNNs the last problem becomes decidable, although the solution length can surpass all computable bounds. If additionally the string is limited to polynomial length, the problem becomes NP-complete and APX-hard. In summary, this shows that approximations and heuristic algorithms are necessary in practical applications of those RNNs.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.