US 12,354,003 B2
Systems and methods for synthetic data generation using a classifier
Anh Truong, Champaign, IL (US); Austin Walters, Savoy, IL (US); and Jeremy Goodsitt, Champaign, IL (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Sep. 14, 2023, as Appl. No. 18/368,318.
Application 18/368,318 is a continuation of application No. 17/520,482, filed on Nov. 5, 2021, granted, now 11,790,229.
Application 17/520,482 is a continuation of application No. 16/686,632, filed on Nov. 18, 2019, granted, now 11,170,298, issued on Nov. 9, 2021.
Prior Publication US 2024/0005153 A1, Jan. 4, 2024
Int. Cl. G06N 3/08 (2023.01); G06N 3/04 (2023.01); G06N 3/044 (2023.01); G06N 5/04 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/044 (2023.01); G06N 5/04 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
obtaining a sequence of elements used to train a recurrent neural network (RNN) for generating synthetic data;
determining, using a classifier trained to identify matches to a token, that at least a portion of the sequence of elements matches the token;
modifying, based on determining that at least the portion of the sequence of elements matches the token, the RNN to use a second vocabulary that includes the token and a first vocabulary; and
repeating, after modifying the RNN and using the second vocabulary, a forward phase of training the RNN for at least the portion of the sequence of elements by reverting the RNN to a prior state the RNN was in when an element, of at least the portion of the sequence of elements, was previously input into the RNN.