TRAINING NEURAL NETWORKS USING NORMALIZED TARGET OUTPUTS

(19)

(11)

EP 3 360 087 B8

(12)	CORRECTED EUROPEAN PATENT SPECIFICATION
	Note: Bibliography reflects the latest situation

(15)	Correction information:
	Corrected version no 1 (W1 B1)

(48)	Corrigendum issued on:
	17.07.2024 Bulletin 2024/29

(45)	Mention of the grant of the patent:
	10.04.2024 Bulletin 2024/15

(21)	Application number: 16809572.7

(22)	Date of filing: 11.11.2016

(51)

International Patent Classification (IPC):

G06N 3/084^(2023.01)
G06N 3/045^(2023.01)

G06N 3/09^(2023.01)

(52)	Cooperative Patent Classification (CPC):
	G06N 3/084; G06N 3/045; G06N 3/09

(86)	International application number:
	PCT/US2016/061663

(87)	International publication number:
	WO 2017/083746 (18.05.2017 Gazette 2017/20)

(54)	TRAINING NEURAL NETWORKS USING NORMALIZED TARGET OUTPUTS TRAINIEREN VON NEURONALEN NETZEN MITTELS NORMALISIERTER ZIELAUSGÄNGE APPRENTISSAGE DE RÉSEAUX NEURONAUX À L'AIDE DE SORTIES CIBLES NORMALISÉES

(84)	Designated Contracting States:
	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

(30)

Priority:

12.11.2015 US 201562254618 P

(43)	Date of publication of application:
	15.08.2018 Bulletin 2018/33

(73)	Proprietor: DeepMind Technologies Limited
	London EC4A 3TW (GB)

(72)	Inventor:
	VAN HASSELT, Hado Philip London N1C 4AG (GB)

(74)	Representative: Marks & Clerk GST
	1 New York Street Manchester M1 4HD Manchester M1 4HD (GB)

(56)

References cited: :

EP-A2- 1 215 627

M. TYGERT ET AL: "Scale-invariant learning and convolutional networks", ARXIV:1506.08230V2, 13 September 2015 (2015-09-13), XP055349551, Retrieved from the Internet <URL:https://arxiv.org/abs/1506.08230v2> [retrieved on 20170222]
Ç. GÜLÇEHRE ET AL: "Knowledge matters: importance of prior information for optimization", ARXIV:1301.4083V6, 13 July 2013 (2013-07-13), XP055266282, Retrieved from the Internet <URL:https://arxiv.org/abs/1301.4083v6> [retrieved on 20170222]
T. VATANEN ET AL: "Pushing stochastic gradient towards second-order methods - backpropagation learning with transformations in nonlinearities", LECTURE NOTES IN COMPUTER SCIENCE, vol. 8226, 3 November 2013 (2013-11-03), pages 442 - 449, XP047044497, DOI: 10.1007/978-3-642-42054-2_55
Y. GANIN, V. LEMPITSKY: "Unsupervised domain adaptation by backpropagation", JMLR WORKSHOP AND CONFERENCE PROCEEDINGS, vol. 37, 1 June 2015 (2015-06-01), pages 1180 - 1189, XP055349552, ISSN: 1938-7228
G. DESJARDINS ET AL: "Natural neural networks", ARXIV:1507.00210V1, 1 July 2015 (2015-07-01), XP055298466, Retrieved from the Internet <URL:https://arxiv.org/abs/1507.00210v1> [retrieved on 20170222]
S. IOFFE, C. SZEGEDY: "Batch normalization: accelerating deep network training by reducing internal covariate shift", ARXIV:1502.03167V3, 2 March 2015 (2015-03-02), XP055266268, Retrieved from the Internet <URL:https://arxiv.org/abs/1502.03167v3> [retrieved on 20170222]

Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).