US 11,868,884 B2
Method and system for providing machine learning service
Jian Gong Deng, Redwood City, CA (US); Ikkjin Ahn, Redwood City, CA (US); Daeseob Lim, Redwood City, CA (US); Bokyung Choi, Redwood City, CA (US); Sechan Oh, Redwood City, CA (US); and William Kanaan, Redwood City, CA (US)
Assigned to MOLOCO, INC., Redwood City, CA (US)
Filed by Moloco, Inc., Redwood City, CA (US)
Filed on Jun. 17, 2020, as Appl. No. 16/904,237.
Claims priority of provisional application 62/893,725, filed on Aug. 29, 2019.
Claims priority of provisional application 62/862,986, filed on Jun. 18, 2019.
Prior Publication US 2020/0401886 A1, Dec. 24, 2020
Int. Cl. G06N 3/08 (2023.01); G06N 3/045 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/045 (2023.01)] 18 Claims
OG exemplary drawing
 
1. A method for providing machine learning model service, the method comprising:
(a) generating, by a first computing system, a first output data using a first machine learning model, wherein the first machine learning model is trained on a first training dataset, and wherein the first output data comprises a prediction generated by the first machine learning model;
(b) transmitting only the first output data to a second computing system, wherein the first training dataset and the first machine learning model are accessible to the first computing system and inaccessible to the second computing system;
(c) joining the first output data generated by the first machine learning model in (a) with a selected set of input features to create an input data to be processed by a second machine learning model, wherein the selected set of input features are accessible to the second computing system and inaccessible to the first computing system and wherein a second training dataset for training the second machine learning model comprises predictions generated by the first machine learning model; and
(d) generating a second output data using the second machine learning model to process the input data, wherein the second machine learning model is inaccessible to the first computing system such that the first machine learning model and the second machine learning model remain isolated and secured in the first computing system and the second computing system respectively.