US 11,836,633 B2
Generating realistic counterfactuals with residual generative adversarial nets
Daniel Alexander Nemirovsky, San Francisco, CA (US); and Nicolas Kevin Thiebaut, San Francisco, CA (US)
Assigned to Vettery, Inc., New York, NY (US)
Filed by Vettery, Inc., New York, NY (US)
Filed on Sep. 8, 2021, as Appl. No. 17/469,339.
Claims priority of provisional application 63/075,578, filed on Sep. 8, 2020.
Prior Publication US 2022/0083871 A1, Mar. 17, 2022
Int. Cl. G06N 3/08 (2023.01); G06N 3/088 (2023.01); G06N 3/045 (2023.01)
CPC G06N 3/088 (2013.01) [G06N 3/045 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
using at least one computer hardware processor to perform:
receiving an input candidate profile associated with a candidate requesting to join a hiring marketplace system;
applying a trained machine learning model to the input candidate profile to obtain a first outcome, the first outcome indicating whether the candidate is permitted to join the hiring marketplace system based on the input candidate profile;
determining whether the first outcome has a value in a set of one or more target values;
when it is determined that the first outcome does not have a value in the set of one or more target values,
generating a counterfactual candidate profile at least in part by using a residual generative adversarial network model including a generator neural network model, the generator neural network model trained using a discriminator neural network model, wherein generating the counterfactual candidate profile comprises:
applying the generator neural network model to the input candidate profile to obtain a corresponding output, wherein:
 the corresponding output indicates changes to be made to one or more values of one or more attributes of the input candidate profile to obtain the counterfactual candidate profile,
 and
 applying the trained machine learning model to the counterfactual candidate profile produces a second outcome having a value in the set of one or more target values, the second outcome indicating that the candidate is permitted to join the hiring marketplace system;
generating feedback based on the counterfactual candidate profile, wherein generating the feedback comprises generating a recommendation that the candidate consider making at least one of the changes indicated in the corresponding output of the generator neural network model;
after providing the recommendation to the candidate, receiving an updated candidate profile including at least one of the changes indicated in the corresponding output of the generator neural network model;
determining whether the candidate is permitted to join the hiring marketplace system at least in part by applying the trained machine learning model to the updated candidate profile; and
permitting the candidate to join the hiring marketplace system in response to determining that the candidate is permitted to join the hiring marketplace system.