US 12,267,189 B2
Neural-network-based receivers
Dani Johannes Korpi, Helsinki (FI); Mikko Aleksi Uusitalo, Helsinki (FI); Janne Matti Juhani Huttunen, Espoo (FI); Leo Mikko Johannes Karkkainen, Helsinki (FI); and Mikko Johannes Honkala, Espoo (FI)
Assigned to Nokia Technologies Oy, Espoo (FI)
Appl. No. 18/021,631
Filed by Nokia Technologies Oy, Espoo (FI)
PCT Filed Aug. 20, 2020, PCT No. PCT/EP2020/073424
§ 371(c)(1), (2) Date Feb. 16, 2023,
PCT Pub. No. WO2022/037785, PCT Pub. Date Feb. 24, 2022.
Prior Publication US 2023/0308317 A1, Sep. 28, 2023
Int. Cl. H04L 25/03 (2006.01); H04B 1/10 (2006.01)
CPC H04L 25/03165 (2013.01) [H04B 1/10 (2013.01)] 13 Claims
OG exemplary drawing
 
1. A node for a telecommunication network, the node comprising a neural-network-based receiver for uplink communications, wherein the node comprises:
at least one processor; and
at least one memory storing instructions that, when executed with the at least one processor, cause the node to:
modify the neural-network-based receiver for uplink communication in the node, to generate a set of modified receiver frameworks defining respective different versions for the neural-network-based receiver;
using the set of modified receiver frameworks, generate respective measures representing bits encoded with a signal received at the node;
calculate a value representing a variance of the respective measures; and
on a basis of the value, determine to select the signal received at the node for use as part of a training data set for the neural-network-based receiver; and
send the signal and a set of corresponding data bits as at least part of the training data set from the node toward a network entity.