Title:
INFORMATION PROCESSING DEVICE AND INFORMATION PROCESSING METHOD
Document Type and Number:
WIPO Patent Application WO/2019/220755
Kind Code:
A1
Abstract:
[Problem] To reduce a computation process load, and perform learning with higher precision. [Solution] Provided is an information processing device provided with a learning unit that, in a neural network quantization function using, as an argument, a parameter for determining a dynamic range, optimizes the parameter for determining a dynamic range by an error backward propagation method and a stochastic gradient descent method. Also, provided is an information processing method comprising causing a processor to, in a neural network quantization function using, as an argument, a parameter for determining a dynamic range, optimize the parameter for determining a dynamic range by an error backward propagation method and a stochastic gradient descent method.
Inventors:
YOSHIYAMA KAZUKI (JP)
UHLICH STEFAN (DE)
CARDINAUX FABIEN (DE)
UHLICH STEFAN (DE)
CARDINAUX FABIEN (DE)
Application Number:
PCT/JP2019/010101
Publication Date:
November 21, 2019
Filing Date:
March 12, 2019
Export Citation:
Assignee:
SONY CORP (JP)
International Classes:
G06N3/08
Other References:
CHOI, JUNGWOOK ET AL.: "PACT: Parameterized Clipping Activation for Quantized Neural Networks", UNDER REVIEW AS A CONFERENCE PAPER AT ICLR 2018, 16 February 2018 (2018-02-16), pages 1 - 17, XP081246007, Retrieved from the Internet [retrieved on 20190422]
TAKEDA, RYU ET AL.: "Acoustic Model Training based on Weight Boundary Model for Discrete Deep Neural Networks", JSAI TECHNICAL REPORT, SIG-CHALLENAE-046-02, 9 November 2016 (2016-11-09), pages 2 - 11, Retrieved from the Internet [retrieved on 20190422]
MIYASHITA, DAISUKE ET AL.: "Convolutional Neural Networks using Logarithmic Data Representation", ARXIV (CORNELL UNIVERSITY, 17 March 2016 (2016-03-17), pages 1 - 10, XP080686928, Retrieved from the Internet [retrieved on 20190422]
LIN, DARRYL D. ET AL.: "Fixed Point Quantization of Deep Convolutional Networks", PROCEEDINGS OF THE 33RD INTERNATIONAL CONFERENCE ON MACHINE LEARNING, vol. 48, 2016, pages 2849 - 2858, XP055561866, Retrieved from the Internet [retrieved on 20190422]
ISHII, JUN ET AL.: "Evaluation of Quantized Bit Width Optimization for Each Neuron for DNN", IPSJ SIG TECHNICAL REPORT, vol. 117, no. 379, 11 January 2018 (2018-01-11), pages 125 - 132
PARK, EUNHYEOK ET AL.: "Weighted-Entropy-based Quantization for Deep Neural Networks", 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, 9 November 2017 (2017-11-09), pages 7197 - 7205, XP033250087, Retrieved from the Internet [retrieved on 20190422]
TAKEDA, RYU ET AL.: "Boundary Contraction Training for Acoustic Models based on Discrete Deep Neural Networks", 15TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION (INTERSPEECH 2014), 14 September 2014 (2014-09-14), pages 1063 - 1067, XP055654496, Retrieved from the Internet [retrieved on 20190422]
TAKEDA, RYU ET AL.: "Acoustic Model Training based on Weight Boundary Model for Discrete Deep Neural Networks", JSAI TECHNICAL REPORT, SIG-CHALLENAE-046-02, 9 November 2016 (2016-11-09), pages 2 - 11, Retrieved from the Internet
MIYASHITA, DAISUKE ET AL.: "Convolutional Neural Networks using Logarithmic Data Representation", ARXIV (CORNELL UNIVERSITY, 17 March 2016 (2016-03-17), pages 1 - 10, XP080686928, Retrieved from the Internet
LIN, DARRYL D. ET AL.: "Fixed Point Quantization of Deep Convolutional Networks", PROCEEDINGS OF THE 33RD INTERNATIONAL CONFERENCE ON MACHINE LEARNING, vol. 48, 2016, pages 2849 - 2858, XP055561866, Retrieved from the Internet
ISHII, JUN ET AL.: "Evaluation of Quantized Bit Width Optimization for Each Neuron for DNN", IPSJ SIG TECHNICAL REPORT, vol. 117, no. 379, 11 January 2018 (2018-01-11), pages 125 - 132
PARK, EUNHYEOK ET AL.: "Weighted-Entropy-based Quantization for Deep Neural Networks", 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, 9 November 2017 (2017-11-09), pages 7197 - 7205, XP033250087, Retrieved from the Internet
TAKEDA, RYU ET AL.: "Boundary Contraction Training for Acoustic Models based on Discrete Deep Neural Networks", 15TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION (INTERSPEECH 2014), 14 September 2014 (2014-09-14), pages 1063 - 1067, XP055654496, Retrieved from the Internet
Attorney, Agent or Firm:
SAKAI INTERNATIONAL PATENT OFFICE (JP)
Download PDF:
Previous Patent: ORIENTATION CONTROL DEVICE
Next Patent: NUMERIC VALUE INPUT DEVICE AND CHARACTER INPUT DEVICE
Next Patent: NUMERIC VALUE INPUT DEVICE AND CHARACTER INPUT DEVICE