| CPC G06N 3/08 (2013.01) [G06F 18/2193 (2023.01); G06F 18/251 (2023.01); G06N 3/044 (2023.01); H04L 67/12 (2013.01)] | 20 Claims |
|
1. A computer-implemented method comprising:
receiving information comprising one or more indications of a plurality of machine learning (ML) models to be used in conjunction with data generated by one or more electronic devices;
analyzing, by a preprocessing adapter executing code using one or more processors, a data structure of an input layer of a first ML model of the plurality of ML models, the analyzing comprising:
generating a plurality of test data elements of a plurality of types;
providing the plurality of test data elements to the first ML model as input data;
receiving a plurality of outputs of the input layer of the first ML model resulting from the providing of the plurality of test data elements to the first ML model;
determining a plurality of loss values based on the plurality of outputs of the input layer of the first ML model;
identifying, as a type of input data accepted by the first ML model, a type of a test data element of the plurality of test data elements that resulted in a lowest loss value of the plurality of loss values; and
determining, by the preprocessing adapter based at least in part on the identifying of the type of the test data element that resulted in the lowest loss value, a first set of one or more preprocessing operations to be performed on the data generated by the one or more electronic devices to yield first preprocessed data of a first type accepted by the first ML model;
performing, by the preprocessing adapter, the first set of preprocessing operations and a second set of preprocessing operations on the data to yield the first preprocessed data of the first type and second preprocessed data of a second type; and
causing the first preprocessed data of the first type to be provided to the first ML model and the second preprocessed data of the second type to be provided to a second ML model.
|