| CPC G06N 20/00 (2019.01) [G06F 16/27 (2019.01); G06N 5/04 (2013.01)] | 19 Claims |
|
1. A method, comprising:
partitioning a machine learning model training dataset into a plurality of training data blocks;
generating a first machine learning model exclusively using a first training data block, of the plurality of training data blocks;
generating a second machine learning model exclusively using a second training data block, of the plurality of training data blocks;
generating an accuracy score for the first machine learning model and an accuracy score for the second machine learning model;
selecting the first training data block and the second training data block for combination with each other based, at least in part, on the accuracy score for the first machine learning model and the accuracy score for the second machine learning model;
generating a third training data block that comprises a combination of the first training data block and the second training data block;
generating a third machine learning model exclusively using the third training data block;
deactivating the first training data block, the second training data block, the first machine learning model, and the second machine learning model; and
activating the third training data block and the third machine learning model
wherein deactivating a training data block makes the training data block unavailable for being selected for generating a combined training data block in an iterative training process until the training data block is reactivated,
wherein deactivating a machine learning model makes the machine learning model unavailable for being selected for performing machine learning predictions or classifications until the machine learning model is reactivated.
|