Finite rank deep kernel learning for robust time series forecasting and regression
AssigneeIntuit Inc.
InventorsSambarta Dasgupta, Sricharan Kumar, Ashok Srivastava
Life of this patent
Estimating the term…
Abstract
Certain aspects of the present disclosure provide techniques for performing finite rank deep kernel learning. In one example, a method for performing finite rank deep kernel learning includes receiving a training dataset; forming a set of embeddings by subjecting the training data set to a deep neural network; forming, from the set of embeddings, a plurality of dot kernels; combining the plurality of dot kernels to form a composite kernel for a Gaussian process; receiving live data from an application; and predicting a plurality of values and a plurality of uncertainties associated with the plurality of values simultaneously using the composite kernel.
Claims20
A finite rank deep kernel learning method, comprising:
- receiving a training dataset;
- training a deep neural network based on the training data set;
- forming a set of embeddings based on the trained deep neural network;
- optimizing an orthogonality of the set of embeddings based on a cost function, the cost function including a penalty term associated with the orthogonality of the set of embeddings;
- forming, from the set of embeddings, a plurality of dot kernels, wherein the plurality of dot kernels consists of a finite number of dot kernels;
- combining the plurality of dot kernels to form a composite kernel for a Gaussian process;
- receiving live data from an application;
- and
- predicting a plurality of values and a plurality of uncertainties associated with the plurality of values simultaneously using the composite kernel.
A system, comprising:
- a memory comprising computer-executable instructions;
- a processor configured to execute the computer-executable instructions and cause the system to perform a finite rank deep kernel learning method, the method comprising:
- receiving a training dataset;
- training a deep neural network based on the training data set;
- forming a set of embeddings based on the trained deep neural network;
- optimizing an orthogonality of the set of embeddings based on a cost function, the cost function including a penalty term associated with orthogonality of the set of embeddings;
- forming, from the set of embeddings, a plurality of dot kernels, wherein the plurality of dot kernels consists of a finite number of dot kernels;
- combining the plurality of dot kernels to form a composite kernel for a Gaussian process;
- receiving live data from an application;
- and
- predicting a plurality of values and a plurality of uncertainties associated with the plurality of values simultaneously using the composite kernel.
A non-transitory computer-readable medium comprising instructions that, when executed by a processor of a processing system, cause the processing system to perform a finite rank deep kernel learning method, the method comprising:
- receiving a training dataset;
- training a deep neural network based on the training data set;
- forming a set of embeddings based on the trained deep neural network;
- optimizing an orthogonality of the set of embeddings based on a cost function, the cost function including a penalty term associated with orthogonality of the set of embeddings;
- forming, from the set of embeddings, a plurality of dot kernels, wherein the plurality of dot kernels consists of a finite number of dot kernels;
- combining the plurality of dot kernels to form a composite kernel for a Gaussian process;
- receiving live data from an application;
- and
- predicting a plurality of values and a plurality of uncertainties associated with the plurality of values simultaneously using the composite kernel.
Dependencies are inferred from the claim text.