Method and apparatus for training a system model with gain constraints
First Claim
1. A system, comprising:
- a model, comprising;
a linear portion; and
a non-linear portion;
wherein the model comprises a representation of a plant or process;
wherein the model is operable to receive an input vector comprising one or more inputs, and compute a predicted output vector comprising one or more outputs corresponding to one or more attributes of the plant or process, and wherein the predicted output vector is usable to manage the plant or process; and
wherein the non-linear portion comprises a function, wherein for each of the one or more inputs;
the function has at most one bend; and
as the input respectively approaches positive and negative infinity, the function asymptotically approaches respective lines of constant slope.
3 Assignments
0 Petitions
Accused Products
Abstract
Method and apparatus for training a system model with gain constraints. A method is disclosed for training a steady-state model, the model having an input and an output and a mapping layer for mapping the input to the output through a stored representation of a system. A training data set is provided having a set of input data u(t) and target output data y(t) representative of the operation of a system. The model is trained with a predetermined training algorithm which is constrained to maintain the sensitivity of the output with respect to the input substantially within user defined constraint bounds by iteratively minimizing an objective function as a function of a data objective and a constraint objective. The data objective has a data fitting learning rate and the constraint objective has constraint learning rate that are varied as a function of the values of the data objective and the constraint objective after selective iterative steps.
-
Citations
24 Claims
-
1. A system, comprising:
-
a model, comprising;
a linear portion; and
a non-linear portion;
wherein the model comprises a representation of a plant or process;
wherein the model is operable to receive an input vector comprising one or more inputs, and compute a predicted output vector comprising one or more outputs corresponding to one or more attributes of the plant or process, and wherein the predicted output vector is usable to manage the plant or process; and
wherein the non-linear portion comprises a function, wherein for each of the one or more inputs;
the function has at most one bend; and
as the input respectively approaches positive and negative infinity, the function asymptotically approaches respective lines of constant slope. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A computer-accessible memory medium that stores program instructions executable by a processor to perform:
-
a model receiving an input vector comprising one or more inputs, wherein the model comprises a representation of a plant or process, wherein the model comprises a linear portion and a non-linear portion, and wherein the non-linear portion comprises a function, and wherein for each of the one or more inputs;
the function has at most one bend; and
as the input respectively approaches positive and negative infinity, the function asymptotically approaches respective lines of constant slope;
computing predicted output corresponding to one or more attributes of the plant or process; and
storing the predicted output, wherein the predicted output is usable to manage the plant or process.
-
-
23. A method, comprising:
-
receiving one or more inputs to a model, wherein the model comprises a representation of a plant or process, wherein the model comprises a linear portion and a non-linear portion, and wherein the non-linear portion comprises a function, wherein the function comprises an integrated sigmoid function, and wherein for each of the one or more inputs;
the function has at most one bend; and
as the input respectively approaches positive and negative infinity, the function asymptotically approaches respective lines of constant slope;
computing predicted output corresponding to one or more attributes of the plant or process; and
storing the predicted output, wherein the predicted output is usable to manage the plant or process.
-
-
24. A system, comprising:
-
a model, comprising;
one or more inputs, operable to receive input data;
a linear portion;
a non-linear portion, wherein the non-linear portion comprises an integrated sigmoid function;
wherein the model comprises a representation of a plant or process; and
wherein the model is operable to receive an input vector comprising one or more inputs, and compute a predicted output vector comprising one or more outputs corresponding to one or more attributes of the plant or process, and wherein the predicted output vector is usable to manage the plant or process.
-
Specification