Title:
DATA PROCESSING SYSTEM AND DATA PROCESSING METHOD
Document Type and Number:
WIPO Patent Application WO/2019/142241
Kind Code:
A1
Abstract:
Provided is a data processing system 100 comprising a learning part for optimizing parameters to be optimized for a neural network on the basis of a comparison between output data outputted by executing, on learning data, a process that accords with the neural network and ideal output data relative to said learning data. With C indicating a first parameter and W indicating a second parameter having a nonnegative value, an activation function f(x) of the neural network is a function for which an output value relative to an input value continuously takes values within the range of C ± W, the output value relative to the input value is uniquely determined, and the graph thereof has point symmetry with respect to a point corresponding to f(x) = C. The learning part optimizes the first parameter and the second parameter as an optimization parameter.
Inventors:
YAGUCHI YOICHI (JP)
Application Number:
PCT/JP2018/001051
Publication Date:
July 25, 2019
Filing Date:
January 16, 2018
Export Citation:
Assignee:
OLYMPUS CORP (JP)
International Classes:
G06N3/08
Foreign References:
| US5271090A | 1993-12-14 |
Other References:
ANONYMOUS: "Artificial Neural Network", WEBARCHIVE.ORG, 8 November 2012 (2012-11-08), XP055625028, Retrieved from the Internet [retrieved on 20180409]
HE, KAIMING ET AL.: "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification", 2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 6 February 2015 (2015-02-06) - 7 December 2015 (2015-12-07), pages 1026 - 1034, XP032866428
KIDA YUSUKE ET AL.: "Voice activity detection based on optimally weighted combination of multiple features", IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, vol. J89-D, 22 August 2006 (2006-08-22) - 9 April 2018 (2018-04-09), pages 1820 - 1828
HE, KAIMING ET AL.: "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification", 2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 6 February 2015 (2015-02-06) - 7 December 2015 (2015-12-07), pages 1026 - 1034, XP032866428
KIDA YUSUKE ET AL.: "Voice activity detection based on optimally weighted combination of multiple features", IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, vol. J89-D, 22 August 2006 (2006-08-22) - 9 April 2018 (2018-04-09), pages 1820 - 1828
Attorney, Agent or Firm:
MORISHITA Sakaki (JP)
Download PDF: