Domain-specific constraints for predictive modeling
A machine learning system that incorporates arbitrary constraints is provided. The machine learning system selects a set of domain-specific constraints from a plurality of sets of domain-specific constraints. The machine learning system selects a set of general functional relationships from a plurality of sets of general functional relationships. The machine learning system maps the selected set of general functional relationships and the selected set of domain-specific constraints to a set of learning transforms. The machine learning system modifies a machine learning specification according to the set of learning transforms, wherein the machine learning specification specifies a model construction, a model setup, and a training objective function. The machine learning system optimizes a machine learning model according to the modified machine learning specification.
1 . A computing device, comprising:
a processor; and
a storage device storing a set of instructions, wherein an execution of the set of instructions by the processor configures the computing device to:
receive a selection of an industry from a plurality of industries via a user interface;
translate the selection of the industry into a selection of a set of domain-specific constraints and a selection of a set of general functional relationships, wherein
the set of domain-specific constraints is selected from a plurality of sets of domain-specific constraints in a constraint mapping library,
the selected set of domain-specific constraints defines relationships specific to the selected industry,
the set of general functional relationships is selected from a plurality of sets of general functional relationships, and
each general functional relationship of the set of general functional relationships defines a transformation applicable to machine learning model construction;
map the selected set of general functional relationships and the selected set of domain-specific constraints to a set of learning transforms;
modify a machine learning specification according to the set of learning transforms, wherein
the machine learning specification specifies a model construction, a model setup, and a training objective function, and
the modification of the machine learning specification comprises:
modifying the model construction by addition of a wide learning model to a non-parametric deep learning model; and
modifying the training objective function by adding an additional loss function for implementing the selected set of domain-specific constraints;
construct a machine learning model based on the modified machine learning specification such that the machine learning model includes the additional loss function and the wide learning model; and
optimize, based on the constructed machine learning model, the machine learning model by determining weights of connections between nodes of the machine learning model.
2 . The computing device of claim 1 , wherein the modification of the machine learning specification according to the set of learning transforms further comprises augmenting a training dataset for the model setup.
3 . The computing device of claim 1 , wherein the model construction comprises the non-parametric deep learning model having one or more intermediate layers.
4 . The computing device of claim 3 , wherein the modification of the machine learning specification according to the set of learning transforms further comprises adding the wide learning model having no intermediate learning layer to the model construction.
5 . The computing device of claim 4 , wherein
the non-parametric deep learning model is unconstrained with respect to the selected set of general functional relationships, and
the wide learning model is constrained with respect to the selected set of general functional relationships.
6 . The computing device of claim 1 , wherein the optimization of the machine learning model comprises performing stochastic gradient descent based on to at least one of the model construction, the model setup, or the training objective function that is modified by the set of learning transforms.
7 . A computer-implemented method, comprising:
receiving a selection of an industry from a plurality of industries via a user interface;
translating the selection of the industry into a selection of a set of domain-specific constraints and a selection of a set of general functional relationships, wherein
the set of domain-specific constraints is selected from a plurality of sets of domain-specific constraints in a constraint mapping library,
the selected set of domain-specific constraints defines relationships specific to the selected industry,
the set of general functional relationships is selected from a plurality of sets of general functional relationships, and
each general functional relationship of the set of general functional relationships defines a transformation applicable to machine learning model construction;
mapping the selected set of general functional relationships and the selected set of domain-specific constraints to a set of learning transforms;
modifying a machine learning specification according to the set of learning transforms, wherein
the machine learning specification specifies a model construction, a model setup, and a training objective function, and
the modifying of the model construction comprises:
modifying the model construction by addition of a wide learning model to a non-parametric deep learning model; and
modifying the training objective function by adding an additional loss function for implementing the selected set of domain-specific constraints;
constructing a machine learning model based on the modified machine learning specification such that the machine learning model includes the additional loss function and the wide learning model; and
optimizing, based on the constructed machine learning model, the machine learning model by determining weights of connections between nodes of the machine learning model.
8 . The computer-implemented method of claim 7 , wherein the modifying of the machine learning specification according to the set of learning transforms further comprises augmenting a training dataset for the model setup.
9 . The computer-implemented method of claim 7 , wherein the model construction comprises the non-parametric deep learning model having one or more intermediate layers.
10 . The computer-implemented method of claim 9 , wherein the modifying of the machine learning specification according to the set of learning transforms further comprises adding the wide learning model having no intermediate learning layer to the model construction.
11 . The computer-implemented method of claim 10 , wherein
the non-parametric deep learning model is unconstrained with respect to the selected set of general functional relationships, and
the wide learning model is constrained with respect to the selected set of general functional relationships.
12 . The computer-implemented method of claim 7 , wherein the optimizing of the machine learning model comprises performing stochastic gradient descent based on at least one of the model construction, the model setup, or the training objective function that is modified by the set of learning transforms.
13 . A computer program product comprising:
one or more non-transitory computer-readable storage devices and program instructions stored on at least one of the one or more non-transitory computer-readable storage devices, the program instructions executable by a processor, cause the processor to perform:
receiving a selection of an industry from a plurality of industries via a user interface;
translating the selection of the industry into a selection of a set of domain-specific constraints and a selection of a set of general functional relationships, wherein
the set of domain-specific constraints is selected from a plurality of sets of domain-specific constraints in a constraint mapping library,
the selected set of domain-specific constraints defines relationships specific to the selected industry,
the set of general functional relationships is selected from a plurality of sets of general functional relationships, and
each general functional relationship of the set of general functional relationships defines a transformation applicable to machine learning model construction;
mapping the selected set of general functional relationships and the selected set of domain-specific constraints to a set of learning transforms;
modifying a machine learning specification according to the set of learning transforms, wherein
the machine learning specification specifies a model construction, a model setup, and a training objective function, and
the modifying of the model construction comprises:
modifying the model construction by addition of a wide learning model to a non-parametric deep learning model; and
modifying the training objective function by adding an additional loss function for implementing the selected set of domain-specific constraints;
constructing a machine learning model based on the modified machine learning specification such that the machine learning model includes the additional loss function and the wide learning model; and
optimizing, based on the constructed machine learning model, the machine learning model by determining weights of connections between nodes of the machine learning model.
14 . The computer program product of claim 13 , wherein the optimizing of the machine learning model comprises performing stochastic gradient descent based on at least one of the model construction, the model setup, or the training objective function that is modified by the set of learning transforms.