US 12,488,232 B2
High-level syntax for priority signaling in neural network compression
Goutham Rangu, Tampere (FI); Hamed Rezazadegan Tavakoli, Espoo (FI); Francesco Cricri, Tampere (FI); Miska Matias Hannuksela, Tampere (FI); and Emre Aksu, Tampere (FI)
Assigned to Nokia Technologies Oy, Espoo (FI)
Filed by Nokia Technologies Oy, Espoo (FI)
Filed on Oct. 1, 2020, as Appl. No. 17/060,658.
Claims priority of provisional application 62/909,495, filed on Oct. 2, 2019.
Prior Publication US 2021/0103813 A1, Apr. 8, 2021
Int. Cl. G06N 3/08 (2023.01); G06N 3/04 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/04 (2013.01)] 33 Claims
OG exemplary drawing
 
1. An apparatus comprising:
at least one processor; and
at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to:
receive with a first device information from a second device, where the information comprises at least one parameter configured to be used for compression of a neural network, where the at least one parameter is in regard to at least one first aspect or task of the neural network, and wherein the information further comprises at least one of the following: a sparsification performance map, or a unification performance map, or a decomposition performance map according to which the compression is to be implemented;
wherein the information comprising the at least one parameter configured to be used for compression of the neural network received with the first device from the second device comprises a mapping of, for each class of a plurality of classes, and for each compression threshold of a plurality of compression thresholds of the neural network, a respective compression threshold of the neural network of the plurality of compression thresholds of the neural network to a respective separate accuracy of a plurality of accuracies of the neural network, wherein the respective separate accuracy of the neural network corresponds to whether samples belong to a class predicted with the neural network for the samples.