Dynamically enhancing supervised learning
AssigneeInternational Business Machines Corporation
InventorsMukundan Sundararajan, Siddharth K. Saraya
Life of this patent
Estimating the term…
Abstract
Embodiments of the present invention provide an approach for dynamically enhancing supervised learning using factor modification based on parsing user input. A user selects an object being displayed incorrectly and provides input as to the reason. The user input is parsed to derive a factor that is contributing to the false outcome. The factor is dynamically altered resulting in a decision path that produces a positive outcome. The change is sent to a model or application owner for final validation and refined training of the machine learning model.
Claims20
A computer-implemented method for dynamically enhancing supervised learning in a machine learning process, comprising the computer-implemented steps of:
- receiving a selection of a data object in an image produced as a result of a machine learning model, wherein the selection indicates that the data object contains a false outcome produced by the machine learning process;
- receiving user input related to a reason for the false outcome;
- parsing the user input to derive a noun and a verb;
- deriving, based on the noun, a factor of the model learning model contributing to the false outcome;
- and
- dynamically altering the factor based on the verb to yield a new decision path within the machine learning model that generates a new display of the data object having a positive outcome.
A system for dynamically enhancing supervised learning in a machine learning process, comprising:
- a memory medium comprising program instructions;
- a bus coupled to the memory medium;
- and
- a processor, for executing the program instructions, coupled to the memory medium that when executing the program instructions causes the system to:
- receive a selection of a data object in an image produced as a result of a machine learning model, wherein the selection indicates that the data object contains a false outcome produced by the machine learning process;
- receive user input related to a reason for the false outcome;
- parse the user input to derive a noun and a verb;
- derive, based on the noun, a factor of the model learning model contributing to the false outcome;
- and
- dynamically alter the factor based on the verb to yield a new decision path within the machine learning model that generates a new display of the data object having a positive outcome.
A computer program product for dynamically enhancing supervised learning in a machine learning process, the computer program product comprising a computer readable storage device, and program instructions stored on the computer readable storage device, to:
- receive a selection of a data object in an image produced as a result of a machine learning model, wherein the selection indicates that the data object contains a false outcome produced by the machine learning process;
- receive user input related to a reason for the false outcome;
- parse the user input to derive a noun and a verb;
- derive, based on the noun, a factor of the model learning model contributing to the false outcome;
- and
- dynamically alter the factor based on the verb to yield a new decision path within the machine learning model that generates a new display of the data object having a positive outcome.
Dependencies are inferred from the claim text.