US 12,066,910 B2
Reinforcement learning based group testing
Lav Raj Varshney, Champaign, IL (US); Yingbo Zhou, Palo Alto, CA (US); Caiming Xiong, Menlo Park, CA (US); and Govardana Sachithanandam Ramachandran, Palo Alto, CA (US)
Assigned to Salesforce, Inc., San Francisco, CA (US)
Filed by Salesforce, Inc., San Francisco, CA (US)
Filed on Oct. 11, 2021, as Appl. No. 17/498,155.
Prior Publication US 2023/0113750 A1, Apr. 13, 2023
Int. Cl. G06F 11/26 (2006.01); G06F 11/36 (2006.01); G06N 20/00 (2019.01)
CPC G06F 11/261 (2013.01) [G06F 11/3696 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computer implemented method for performing group testing using reinforcement learning, the method comprising:
receiving a population of items including a subset of defective items;
receiving a request to identify defective items from the population of items;
initializing a reinforcement learning based agent;
initializing a set of pools of items from the population of items;
performing group testing of the population of items using reinforcement learning, comprising:
performing testing on a current pool of items to identify defective items;
determining, by the reinforcement learning based agent, a reward based on results of the testing; and
adjusting, by the reinforcement learning based agent, the current pool of items based on the reward; and
transmitting a result of a final testing based on a final pool of items after the adjusting, the result of the final testing identifying one or more defective items from the population of items.