US 11,836,220 B2
Updating of statistical sets for decentralized distributed training of a machine learning model
Xiaodong Cui, Chappaqua, NY (US); Wei Zhang, Elmsford, NY (US); Mingrui Liu, College Station, TX (US); Abdullah Kayi, Elmsford, NY (US); Youssef Mroueh, New York, NY (US); and Alper Buyuktosunoglu, White Plains, NY (US)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Mar. 1, 2023, as Appl. No. 18/176,765.
Application 18/176,765 is a continuation of application No. 17/159,710, filed on Jan. 27, 2021, granted, now 11,636,280.
Prior Publication US 2023/0205843 A1, Jun. 29, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 18/214 (2023.01); G06F 15/173 (2006.01); G06N 20/00 (2019.01); G06N 3/08 (2023.01); G06F 18/20 (2023.01)
CPC G06F 18/214 (2023.01) [G06F 15/17375 (2013.01); G06F 18/285 (2023.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system, comprising:
a memory that stores computer executable components; and
a processor that executes the computer executable components stored in the memory, wherein the computer executable components comprise:
a statistic set updating component that updates a statistical set employed for training a machine learning node of a group of machine learning nodes distributed on a network, comprising iteratively performing until a convergence criterion is satisfied:
select, using a randomization pattern, an additional statistical set from respective statistical sets employed for training the group of machine learning nodes, wherein the additional statistical set is other than the statistical set; and
update the statistical set associated with the machine learning node by averaging the statistical set with the additional statistical set.