Computational power of neural networks: a Kolmogorov complexity characterization

The computational power of neural networks depends on properties of the real numbers used as weights. We focus on networks restricted to compute in polynomial time, operating on boolean inputs. Previous work has demonstrated that their computational power happens to coincide with the complexity clas...

Descripción completa

Detalles Bibliográficos
Autores: Balcázar Navarro, José Luis|||0000-0003-4248-4528, Gavaldà Mestre, Ricard|||0000-0003-4736-7179, Siegelmann, H.
Tipo de recurso: informe técnico
Fecha de publicación:1993
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/82120
Acceso en línea:https://hdl.handle.net/2117/82120
Access Level:acceso abierto
Palabra clave:Neural networks
Turing Machines
Kolmogorov complexity
Àrees temàtiques de la UPC::Informàtica
Descripción
Sumario:The computational power of neural networks depends on properties of the real numbers used as weights. We focus on networks restricted to compute in polynomial time, operating on boolean inputs. Previous work has demonstrated that their computational power happens to coincide with the complexity classes P and P/poly, respectively, for networks with rational and arbitrary real weights. Here we prove that the crucial concept that characterizes this computational power is the Kolmogorov complexity of the weights, in the sense that, for each bound on this complexity, the networks can solve exactly the problems in a related nonuniform complexity class located between P and P/poly. By proving that the family of such nonuniform classes is infinite, we show that neural networks can be classified into an infinite hierarchy of different computing capabilities.