Back to results
US 12,608,445 B2Granted
Assessing machine learning bias using model training metadata
AssigneeCisco Technology, Inc.
InventorsChiara Troiani, Aviva Vaknin, Frank Michaud
Life of this patent
Estimating the term…
Abstract
In one embodiment, a device receives a request for a machine learning model to make an inference about input data included in the request. The device retrieves metadata regarding training data used to train the machine learning model from a ledger associated with the machine learning model. The device assesses bias of the machine learning model by comparing the input data in the request to the metadata from the ledger. The device provides an indication of the bias of the machine learning model for display.
Claims20
A method comprising:
- receiving, at a device, a request for a machine learning model to make an inference about input data included in the request;
- retrieving, by the device, metadata regarding training data used to train the machine learning model from a ledger associated with the machine learning model;
- comparing, by the device, the input data in the request to the metadata regarding the training data to quantify an amount of deviation between the input data from the training data of the machine learning model when the input data differs from the training data;
- determining, by the device, that the amount of deviation indicates that the training data used to train the machine learning model is non-representative of the input data such that the machine learning model has been trained to be biased with respect to making the inference about the input data in the request;
- and
- providing, by the device, an indication of the bias of the machine learning model for display based on the deviation of the input data from the training data of the machine learning model.
3 more elements
An apparatus, comprising:
- one or more network interfaces;
- a processor coupled to the one or more network interfaces and configured to execute one or more processes;
- and
- a memory configured to store a process that is executable by the processor, the process when executed configured to:
- receive a request for a machine learning model to make an inference about input data included in the request;
- retrieve metadata regarding training data used to train the machine learning model from a ledger associated with the machine learning model;
- compare the input data in the request to the metadata regarding the training data to quantify an amount of deviation between the input data from the training data of the machine learning model when the input data differs from the training data;
- determine that the amount of deviation indicates that the training data used to train the machine learning model is non-representative of the input data such that the machine learning model has been trained to be biased with respect to making the inference about the input data in the request;
- and
- provide an indication of the bias of the machine learning model for display based on the deviation of the input data from the training data of the machine learning model.
7 more elements
A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute a process comprising:
- receiving, at the device, a request for a machine learning model to make an inference about input data included in the request;
- retrieving, by the device, metadata regarding training data used to train the machine learning model from a ledger associated with the machine learning model;
- comparing, by the device, the input data in the request to the metadata regarding the training data to quantify an amount of deviation between the input data from the training data of the machine learning model when the input data differs from the training data;
- determining, by the device, that the amount of deviation indicates that the training data used to train the machine learning model is non-representative of the input data such that the machine learning model has been trained to be biased with respect to making the inference about the input data in the request;
- and
- providing, by the device, an indication of the bias of the machine learning model for display based on the deviation of the input data from the training data of the machine learning model.
3 more elements
Dependencies are inferred from the claim text.