Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
INFORMATION PROCESSING DEVICE AND INFORMATION PROCESSING METHOD
Document Type and Number:
WIPO Patent Application WO/2019/064774
Kind Code:
A1
Abstract:
[Problem] To further reduce processing burden relating to an inner product operation and to secure a quantization granularity of a weight coefficient. [Solution] Provided is an information processing device provided with a product-sum operation circuit for executing a product-sum operation on the basis of a plurality of input values and a plurality of weight coefficients which respectively correspond to the input values and are quantized by a power expression. Exponents of the quantized weight coefficients are expressed by fractions having a predetermined divisor as the denominator, and the product-sum operation circuit performs a product-sum operation using different adding multipliers on the basis of the remainder determined from the divisor.

Inventors:
HIROI, Toshiyuki (1-7-1 Konan, Minato-k, Tokyo 75, 〒1080075, JP)
NAKAMURA, Akira (1-7-1 Konan, Minato-k, Tokyo 75, 〒1080075, JP)
YAMAMOTO, Makiko (1-7-1 Konan, Minato-k, Tokyo 75, 〒1080075, JP)
IKEGAYA, Ryoji (1-7-1 Konan, Minato-k, Tokyo 75, 〒1080075, JP)
Application Number:
JP2018/024923
Publication Date:
April 04, 2019
Filing Date:
June 29, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
SONY CORPORATION (1-7-1, Konan Minato-k, Tokyo 75, 〒1080075, JP)
International Classes:
G06F17/10; G06N3/02
Foreign References:
JPH0451384A1992-02-19
Other References:
MIYASHITA, DAISUKE ET AL.: "Convolutional Neural Networks using Logarithmic Data Representation", 17 March 2016 (2016-03-17), XP080686928, Retrieved from the Internet [retrieved on 20180822]
HUBARA, ITAY ET AL.: "Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations", 22 September 2016 (2016-09-22), XP080813052, Retrieved from the Internet [retrieved on 20180822]
TANG, CHUAN ZHANG ET AL.: "Multilayer Feedforward Neural Networks with Single Powers-of-Two Weights", IEEE TRANSACTIONS ON SIGNAL PROCESSING, vol. 41, no. 8, August 1993 (1993-08-01), pages 2724 - 2727, XP055503286, Retrieved from the Internet [retrieved on 20180822]
MATTHIEU COURBARIAUX ET AL.: "BinaryConnect: Training Deep Neural Networks with binary weights during propagations", ARXIV, 11 November 2015 (2015-11-11), Retrieved from the Internet
DAISUKE MIYASHITA ET AL.: "Convolutional Neural Networks using Logarithmic Data Representation", ARXIV, 3 March 2016 (2016-03-03), Retrieved from the Internet
Attorney, Agent or Firm:
KAMEYA, Yoshiaki et al. (HAZUKI INTERNATIONAL YOTSUYA, Daiichi Tomizawa Building 3-1-3, Yotsuya, Shinjuku-k, Tokyo 04, 〒1600004, JP)
Download PDF: