US 11,989,268 B2
Dummy class framework for continual supervised learning applications
Vinay Uday Prabhu, Redwood City, CA (US); and John C. Whaley, Redwood City, CA (US)
Assigned to UNIFYID, San Francisco, CA (US)
Filed by UnifyID, San Francisco, CA (US)
Filed on Feb. 11, 2021, as Appl. No. 17/174,238.
Claims priority of provisional application 62/975,670, filed on Feb. 12, 2020.
Prior Publication US 2021/0248215 A1, Aug. 12, 2021
Int. Cl. G06F 21/31 (2013.01); G06F 17/18 (2006.01); G06N 3/08 (2023.01)
CPC G06F 21/31 (2013.01) [G06F 17/18 (2013.01); G06N 3/08 (2013.01)] 23 Claims
OG exemplary drawing
 
1. A method comprising:
during a training mode,
training a multilayer neural network model to authenticate one or more initial users based at least in part on an initial training data set comprising sensor data from one or more electronic devices associated with the one or more initial users, wherein the one or more initial users comprises a first user of a first electronic device, and
updating the trained multilayer neural network model based at least in part on an additional training data set which includes sensor data from one or more additional electronic devices associated with one or more additional users to enable authentication of both the one or more initial users and the one or more additional users, wherein the trained multilayer neural network includes at least one weight and wherein the updating based at least in part on the additional training data updates at least a first weight in the at least one weight; and
during a subsequent mode,
authenticating, via the updated trained multilayer neural network model, the first user of the first electronic device based at least in part on sensor data contemporaneously received from the first electronic device.