| CPC G06N 20/00 (2019.01) [G06F 18/26 (2023.01); G06F 18/285 (2023.01); G06F 21/57 (2013.01)] | 20 Claims |
|
1. A computer-implemented method for updating a machine-learning-based prediction model with preserved privacy, at least a portion of the method being performed by a computing device comprising at least one processor, the method comprising:
receiving a plurality of training data sets from a plurality of computing devices, with each data set including incidents of a computing event, for training a machine-learning-based global prediction model for predicting future incidents of the computing event;
creating, by training the global prediction model using the plurality of training data sets, a plurality of intermediate prediction models of the global prediction model, wherein each intermediate prediction model corresponds to a distinct precursor state associated with fully training the global prediction model, and wherein each distinct precursor state corresponds to a different training midpoint with a different amount of bias to the plurality of training data sets;
receiving, from a computing node, an expected size of a local training data set that is local to the computing node and that is separate from the plurality of training data sets, wherein the expected size of the local training data set is indicative of an influence of the local training data set for training;
selecting an intermediate prediction model from the plurality of intermediate prediction models based on correlating the expected size to the amount of bias of the selected intermediate prediction model to the plurality of training data sets to allow the influence of the local training data set to offset the bias of the selected intermediate model; and
providing the selected intermediate prediction model to the computing node to enable the computing node to fully train a local prediction model using both the intermediate prediction model and the local training data set.
|