US 12,423,957 B2
Systems and methods for a lightweight pattern-aware generative adversarial network
Chandrajit Pal, Khandi (IN); Manmohan Tripathi, Hyderabad (IN); and Govardhan Mattela, Hyderabad (IN)
Assigned to Ceremorphic, Inc., San Jose, CA (US)
Filed by Ceremorphic, Inc., San Jose, CA (US)
Filed on Nov. 5, 2021, as Appl. No. 17/519,581.
Prior Publication US 2023/0146468 A1, May 11, 2023
Int. Cl. G06V 10/774 (2022.01); G06N 3/045 (2023.01); G06V 10/40 (2022.01)
CPC G06V 10/7747 (2022.01) [G06N 3/045 (2023.01); G06V 10/513 (2022.01)] 21 Claims
OG exemplary drawing
 
1. A computer-implemented method includes training at least a generative adversarial network, the method operable on one or more processors, the method comprising:
receiving a set of training data comprising elements of multi-valued data where each element of multi-valued data is independent and unrelated to other elements of multi-valued data;
applying pattern extraction to the set of training data to extract one or more feature embeddings representing one or more features of the training data, the feature embeddings comprising subsets of the multi-valued data;
attenuating the one or more feature embeddings to create one or more attenuated feature embeddings;
providing the one or more attenuated embeddings to a generator of the generative adversarial network as a condition to at least partly control the generator in generating synthetic data, the providing being performed automatically and dynamically during training of the generator;
with the generator, generating synthetic data based at least in part on the attenuated embeddings; and
a discriminator performing a discrimination of the synthetic data and the training data comprising a convolution of the synthetic data with the training data to generate a probability indicating whether to use or discard the synthetic data.