US 12,608,601 B2 · Statistical confidence metric for reconstructive anomaly detection models

Abstract

Approaches herein relate to reconstructive models such as an autoencoder for anomaly detection. Herein are machine learning techniques that measure inference confidence based on reconstruction error trends. In an embodiment, a computer hosts a reconstructive model that encodes and decodes features. Based on that decoding, the following are automatically calculated: a respective reconstruction error of each feature, a respective moving average of reconstruction errors of each feature, an average of the moving averages of the reconstruction errors of all features, a standard deviation of the moving averages of the reconstruction errors of all features, and a confidence of decoding the features that is based on a ratio of the average of the moving averages of the reconstruction errors to the standard deviation of the moving averages of the reconstruction errors. The computer detects and indicates that a threshold exceeds the confidence of decoding, which may cause important automatic reactions herein.

Claims18

Dependent claims

A method comprising:

a reconstructive machine learning model encoding and decoding a plurality of features in a new dataset;
automatically calculating based on said decoding the plurality of features:
a respective reconstruction error of each feature of the plurality of features,
a respective moving average of said reconstruction error of said each feature,
an average of said moving averages of said reconstruction errors of the plurality of features,
a standard deviation of said moving averages of said reconstruction errors of the plurality of features, and
a confidence of said decoding the plurality of features that is based on a ratio of said average of said moving averages of said reconstruction errors to said standard deviation of said moving averages of said reconstruction errors;
detecting that a retrain threshold exceeds said confidence of said decoding the plurality of features;
and
retraining, in response to said detecting, the reconstructive machine learning model with the new dataset;
wherein the method is performed by one or more computers.

8 more elements

Dependencies are inferred from the claim text.

Statistical confidence metric for reconstructive anomaly detection models

Life of this patent

Statistical confidence metric for reconstructive anomaly detection models

Life of this patent

Abstract

Claims18