Cross-model format comparison
AssigneeHiddenLayer, Inc.
InventorsKasimir Ralph Peter Schulz, David Beveridge
Life of this patent
Estimating the term…
Abstract
A machine learning model representation is obtained from a model source and information characterizing the layers of the model representation is extracted to result in extracted model information. This extracted model information can be compared to information characterizing one or more known (i.e., previously characterized) machine learning models in order to determine whether there is a match based on layer information. A match can, in some cases, be used to determine an identity of the underlying machine learning model for the model representation. Information regarding the comparison (i.e., the model matching determination) can be provided to a consuming application or process. Related apparatus, systems, techniques and articles are also described.
Claims25
A method for confirming an identity of a machine learning model, the method being implemented by at least one computing device and comprising:
- accessing a first cloud-based machine learning model repository to obtain first data characterizing a first machine learning model;
- extracting first information characterizing weight layers from the first machine learning model including a number of layers and identifiers of such layers;
- accessing a second cloud-based machine learning model repository to obtain second data characterizing a second machine learning model, the second machine learning model being known to operate in a desired manner;
- extracting second information characterizing weight layers from the second machine learning model a number of layers and identifiers of such layers;
- determining, by comparing the extracted first information to the extracted second information, whether the first machine learning model is a same model as the second machine learning model if the respective number of layers and the identifiers of the layers match;
- and
- providing data indicating whether the models store a same or different model to a consuming application or process.
A method for confirming an identity of a machine learning model, the method being implemented by at least one computing device and comprising:
- accessing a first cloud-based model repository to obtain first data characterizing a first machine learning model;
- extracting first information characterizing weight layers from the first machine learning model including a number of layers, identifiers of the layers, weight values for the layers, and bias values for the layers;
- accessing a second cloud-based model repository to obtain a second machine learning model which is of a different format and/or has different metadata than the first machine learning model, the second machine learning model being known to operate in a desired manner;
- extracting second information characterizing weight layers from the second machine learning model including a number of layers, identifiers of the layers, weight values for the layers, and bias values for the layers;
- first comparing the number of layers from the extracted first information to the number of layers from the extracted second information to detect a match;
- second comparing, if a match is detected in the first comparing, the identifiers of the layers from the extracted first information to the identifiers from the extracted second information to detect a match;
- third comparing, if a match is detected in the second comparing, the weight values from the extracted first information to the weight values from the extracted second information to detect a match;
- fourth comparing, if a match is detected in the third comparing, the bias values from the extracted first information to the bias values from the extracted second information to detect a match;
- determining that the first machine learning model is a same model as the second machine learning model if each of the first comparing, the second comparing, the third comparing, and fourth comparing detect a match;
- determining that the first machine learning model does not store a same model as the second machine learning model if one of the first comparing, the second comparing, the third comparing, and the fourth comparing do not detect a match;
- and
- providing data indicating whether the first machine learning model and the second machine learning model are the same or different to a consuming application or process.
A method for confirming an identity of a machine learning model, the method being implemented by at least one computing device and comprising:
- accessing a cloud-based repository to obtain first data characterizing a first machine learning model;
- extracting first information characterizing weight layers from the first machine learning model including a number of layers and identifiers of such layers;
- determining, by comparing the extracted first information to extracted second information associated with a second machine learning model, whether the first machine learning model stores a same model as the second machine learning model, the extracted second information characterizing weight layers from the second machine learning model including a number of layers and identifiers of such layers, the determining being based on whether the respective number of layers and the identifiers of the layers from the extracted first information match and the extracted second information, the second machine learning model being known to operate in a desired manner;
- and
- providing data indicating whether the models store a same or different model to a consuming application or process.
Dependencies are inferred from the claim text.