US 11,989,262 B2
Unsupervised domain adaptation with neural networks
David Acuna Marrero, Toronto (CA); Guojun Zhang, Waterloo (CA); Marc Law, Ontario (CA); and Sanja Fidler, Toronto (CA)
Assigned to Nvidia Corporation, Santa Clara, CA (US)
Filed by Nvidia Corporation, Santa Clara, CA (US)
Filed on Apr. 9, 2021, as Appl. No. 17/226,534.
Claims priority of provisional application 63/086,544, filed on Oct. 1, 2020.
Prior Publication US 2022/0108134 A1, Apr. 7, 2022
Int. Cl. G06F 18/214 (2023.01); G06F 18/21 (2023.01); G06F 18/241 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06V 10/40 (2022.01)
CPC G06F 18/2148 (2023.01) [G06F 18/217 (2023.01); G06F 18/241 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06V 10/40 (2022.01)] 14 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
encoding, into a latent space using a first neural network, a set of features representative of an input;
classifying the input, based at least in part upon the encoded features, using a second neural network;
inferring a domain, corresponding to the input, based at least in part upon the encoded features and using a third neural network; and
adjusting one or more network parameters, for at least the first neural network, until the second neural network accurately classifies the input based on the features encoded by the first neural network, wherein the adjusting the one or more network parameters prevents the third neural network from being able to infer the domain with at least a minimum amount of certainty, and the adjusting is based on a common loss function including terms for each of the first neural network, the second neural network, and the third neural network.