Approximate Homomorphisms and Convergent Representations in Transducers
Not provided
Abstract
The paper investigates the stability of minimal representations of controlled stochastic processes, particularly transducers, under perturbations.
Reality Card
The study proves that minimal linear transducers implementing interfaces close to a finite-rank interface have an approximate homomorphism to the minimal implementation of that interface, with error linear in the perturbation size.
For every finite-rank interface, all minimal linear transducers have an approximate homomorphism with error linear in the perturbation size.
The results depend on mild hypotheses regarding the indistinguishability of belief states, which may limit generalizability.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.