Differentially private open-classification machine learning model training with adaptive clipping

One or more non-transitory computer-readable program instructions that, when executed on or across one or more processors, cause the one or more processors to:

• train an open-classification machine learning model using a set of training data comprising a set of data not subject to privacy constraints and a set of data subject to privacy constraints, wherein said training of the open-classification machine learning model is performed via a plurality of partial training iterations, wherein an individual iteration of partial training uses one or more individual members of the set of data not subject to privacy constraints and one or more individual members of the set of data subject to privacy constraints;
• and
• adjust an influence value for the plurality of partial training iterations, wherein said adjustment of the influence value is limited to a threshold level based on a prior influence value of a prior iteration of partial training of said training of the open-classification machine learning model, and wherein the influence value:
• is based on one or more of the individual members of the set of data not subject to privacy constraints the individual iteration of partial training is using, and
• limits an effect of the one or more of the individual members of the set of data subject to privacy constraints on a weight for a loss amount of the individual iteration of partial training.

A method, comprising,

• training an open-classification machine learning model using a set of training data comprising a set of data not subject to privacy constraints and a set of data subject to privacy constraints, wherein said training of the open-classification machine learning model is performed via a plurality of partial training iterations, wherein an individual iteration of partial training uses one or more individual members of the set of data not subject to privacy constraints and one or more individual members of the set of data subject to privacy constraints;
• and
• adjusting an influence value for the plurality of partial training iterations, wherein said adjusting of the influence value is limited to a threshold level based on a prior influence value of a prior iteration of partial training of said training of the open-classification machine learning model, and wherein the influence value:
• is based on one or more of the individual members of the set of data not subject to privacy constraints the individual iteration of partial training is using, and
• limits an effect of the one or more of the individual members of the set of data subject to privacy constraints on a weight for a loss amount of the individual iteration of partial training.