Probabilistic Bounds on Complexity of Networks Computing Binary Classification Tasks

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F67985807%3A_____%2F18%3A00493926" target="_blank" >RIV/67985807:_____/18:00493926 - isvavai.cz</a>

Result on the web

<a href="http://ceur-ws.org/Vol-2203/86.pdf" target="_blank" >http://ceur-ws.org/Vol-2203/86.pdf</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Probabilistic Bounds on Complexity of Networks Computing Binary Classification Tasks

Original language description

Complexity of feedforward networks computing binary classification tasks is investigated. To deal with unmanageably large number of these tasks on domains of even moderate sizes, a probabilistic model characterizing relevance of the classification tasks is introduced. Approximate measures of sparsity of networks computing randomly chosen functions are studied in terms of variational norms tailored to dictionaries of computational units. Probabilistic lower bounds on these norms are derived using the Chernoff-Hoeffding Bound on sums of independent random variables, which need not be identically distributed. Consequences of the probabilistic results on the choice of dictionaries of computational units are discussed.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

<a href="/en/project/GA18-23827S" target="_blank" >GA18-23827S: Capabilities and limitations of shallow and deep networks</a><br>

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year

2018

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

ITAT 2018: Information Technologies – Applications and Theory. Proceedings of the 18th conference ITAT 2018

ISBN

—

ISSN

1613-0073

e-ISSN

—

Number of pages

6

Pages from-to

86-91

Publisher name

Technical University & CreateSpace Independent Publishing Platform

Place of publication

Aachen

Event location

Plejsy

Event date

Sep 21, 2018

Type of event by nationality

EUR - Evropská akce

UT code for WoS article

—