Reference library
On the Computational Power of Neural Nets
Paper summary
Proves that finite recurrent networks with rational coefficients and suitable sigmoidal dynamics can simulate Turing machines. Explicit constructions and complexity analysis show computational power without higher-order multiplicative networks.
Relevance to the project
Supports the expressiveness of a simple recurrent substrate. Distinguish rational parameters with unbounded state precision from the project's finite floating-point representation.
Computability constructions do not imply efficient discovery or robust finite-precision execution.
Source record
- Reference ID
- doi-10-1006-jcss-1995-1013
- Archived version
- Author-hosted scan of published JCSS 1995 article, pages 132–150
- Summary provenance
- reference-library/papers/doi-10-1006-jcss-1995-1013/metadata.json · summary
- Review status
- initial orientation summary; not a critical full-paper review
- Reading scope
- Visual inspection of scan PDF page 1 (abstract and opening text) and last page (references); no searchable layer. Initial orientation only, not a full proof or methods review.