US 12,229,642 B2
Efficient duplicate detection for machine learning data sets
Leo Parker Dirac, Seattle, WA (US); and Aleksandr Mikhaylovich Ingerman, Seattle, WA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Mar. 26, 2021, as Appl. No. 17/214,047.
Application 17/214,047 is a continuation of application No. 14/569,458, filed on Dec. 12, 2014, granted, now 10,963,810.
Application 14/569,458 is a continuation in part of application No. 14/319,902, filed on Jun. 30, 2014, granted, now 10,102,480, issued on Oct. 15, 2018.
Prior Publication US 2021/0374610 A1, Dec. 2, 2021
Int. Cl. G06N 20/00 (2019.01)
CPC G06N 20/00 (2019.01) 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
obtaining, via one or more programmatic interfaces at a service of a cloud computing environment, a request to tune one or more hyper-parameters of a machine learning model which performs one or more machine learning tasks, wherein tuning the one or more hyper-parameters of the machine learning model affect the result of the one or more machine learning tasks;
based on the one or more hyper-parameters, selecting by the service, respective ranges of values applicable for individual ones of the one or more hyper-parameters to determine respective candidate values for the one or more hyper-parameters from the respective ranges of values that optimize the one or more hyper-parameters, wherein the respective ranges of values are selected by the service based, at least in part, on:
one or more respective ranges of values of the service for one or more previously obtained requests to tune one or more hyper-parameters of one or more machine learning models;
performing one or more tuning operations in which the respective ranges of values for individual ones of the one or more hyper-parameters are utilized for the machine learning model;
evaluating respective results of the one or more tuning operations to generate the respective candidate values for the one or more hyper-parameters according to the evaluation based, at least in part, on a comparison of results for the one or more machine learning tasks that selects the respective candidate values out of the ranges of values for the individual ones of the one or more hyper-parameters according to result similarity; and
providing, via the one or more programmatic interfaces, the respective candidate values for the one or more hyper-parameters.