Global Patent Index - EP 3449424 A1

EP 3449424 A1 20190306 - DEVICE AND METHOD FOR DISTRIBUTING CONVOLUTIONAL DATA OF A CONVOLUTIONAL NEURAL NETWORK

Title (en)

DEVICE AND METHOD FOR DISTRIBUTING CONVOLUTIONAL DATA OF A CONVOLUTIONAL NEURAL NETWORK

Title (de)

VORRICHTUNG UND VERFAHREN ZUR VERTEILUNG VON FALTUNGSDATEN EINES NEURONALEN FALTUNGSNETZWERKS

Title (fr)

DISPOSITIF ET PROCEDE DE DISTRIBUTION DE DONNEES DE CONVOLUTION D'UN RESEAU DE NEURONES CONVOLUTIONNEL

Publication

EP 3449424 A1 20190306 (FR)

Application

EP 17720462 A 20170427

Priority

  • FR 1653744 A 20160427
  • EP 2017060017 W 20170427

Abstract (en)

[origin: WO2017186830A1] The invention proposes a device for distributing convolution coefficients of at least one convolution kernel of a convolutional neural network, carried by an input bus (201), to a set of processing units in a calculator based on a convolutional neural network architecture. The device comprises at least one permutation network (30) controlled by at least one control unit (32), the permutation network comprising a set of permutation units (30) arranged to perform circular offsets of at least a portion of the input bus. For each convolution kernel, each control unit is configured to dynamically control at least some of the permutation units of the permutation networks (30) in response to an input event applied to the convolution kernel and at least one parameter representing the maximum size of the convolution kernels.

IPC 8 full level

G06N 3/04 (2006.01); G06N 3/063 (2006.01)

CPC (source: EP US)

G06N 3/045 (2023.01 - EP); G06N 3/049 (2013.01 - EP US); G06N 3/063 (2013.01 - EP US); G06N 3/08 (2013.01 - US)

Designated contracting state (EPC)

AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

Designated extension state (EPC)

BA ME

DOCDB simple family (publication)

WO 2017186830 A1 20171102; EP 3449424 A1 20190306; FR 3050846 A1 20171103; FR 3050846 B1 20190503; US 11423296 B2 20220823; US 2019156201 A1 20190523

DOCDB simple family (application)

EP 2017060017 W 20170427; EP 17720462 A 20170427; FR 1653744 A 20160427; US 201716095923 A 20170427