	US 11,669,717 B2
	Method and device for completing social network using artificial neural network
Won Yong Shin, Seoul (KR); and Tien Cong Tran, Yongin-si (KR)
Assigned to INDUSTRY-ACADEMIC COOPERATION FOUNDATION YONSEI UNIVERSITY, Seoul (KR)
Filed by INDUSTRY-ACADEMIC COOPERATION FOUNDATION, YONSEI UNIVERSITY, Seoul (KR)
Filed on Jun. 12, 2020, as Appl. No. 16/900,609.
Claims priority of application No. 10-2019-0126749 (KR), filed on Oct. 14, 2019.
Prior Publication US 2021/0110242 A1, Apr. 15, 2021
Int. Cl. G06N 3/047 (2023.01); G06F 16/901 (2019.01); G06N 5/04 (2023.01); G06N 3/08 (2023.01)

CPC G06N 3/047 (2023.01) [G06F 16/9024 (2019.01); G06N 3/08 (2013.01); G06N 5/04 (2013.01)]

10 Claims

1. A network completion device comprising:

a neural network unit configured to receive input of a target network having unrevealed missing nodes, infer connections of the missing nodes by way of a neural network, and output a plurality of candidate complete networks according to various node sequences; and

a selection unit configured to select one of the plurality of candidate complete networks outputted by the neural network unit,

wherein the neural network unit outputs the plurality of candidate complete networks by using weights of a graph-generating neural network, wherein the graph-generating neural network has a learned graph structure of reference networks having the same attributes as attributes of the target network,

the selection unit uses a connection probability vector obtained from the graph-generating neural network to select a candidate complete network probabilistically having a structure closest to a structure of the target network based on the connection probability vector, and

the selection unit selects one of the candidate complete networks based on an equation shown below:

where G represents the candidate complete networks, G₀represents the target network, θ represents the connection probability vector, and Ĝ represents the selected candidate complete network.