Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
TRAINING DEVICE, ESTIMATION DEVICE, TRAINING METHOD, ESTIMATION METHOD, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2020/217620
Kind Code:
A1
Abstract:
An objective of the present invention is to train a high-precision graph generation algorithm. The training device comprises one or a plurality of memories and one or a plurality of processors. The one or the plurality of processors train a first converter for transforming a first feature value relating to a node of a graph and a second feature value relating to a structure of the graph into a first latent value by a transformation whereby an inverse transformation can be defined, and a second converter which transforms the second feature value into a second latent value by a transformation whereby an inverse transformation can be defined, the first converter and the second converter being trained on the basis of the first latent value and the second latent value.

Inventors:
PITUWALAKANKANAMGE KAUSHALYA MADHAWA (JP)
NAKAGO KOSUKE (JP)
ISHIGURO KATSUHIKO (JP)
Application Number:
PCT/JP2020/003052
Publication Date:
October 29, 2020
Filing Date:
January 28, 2020
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
PREFERRED NETWORKS INC (JP)
International Classes:
G06N20/00
Other References:
JEPSEN, TOBIAS SKOVGAARD: "How to do Deep Learning on Graphs with Graph Convolutional Networks", PART 1: A HIGH-LEVEL INTRODUCTION TO GRAPH CONVOLUTIONAL NETWORKS, 18 September 2018 (2018-09-18), XP055689132, Retrieved from the Internet [retrieved on 20200406]
HAMON, RONAN ET AL.: "From graphs to signals and back: Identification of network structures using spectral analysis", ARXIV:1502.04697V3, 10 June 2016 (2016-06-10), pages 1 - 19, XP055758436, Retrieved from the Internet [retrieved on 20200406]
Attorney, Agent or Firm:
NAKAMURA Yukitaka et al. (JP)
Download PDF: