Accurate and query-efficient model agnostic explanations
One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to a process for providing an explanation result for an analytical model. A system can comprise a memory that stores computer executable components, and a processor that executes the computer executable components stored in the memory, wherein the computer executable components can comprise an uncertainty component that determines an uncertainty score for a distribution of samples that neighbor a selected input to an analytical model, a sampling component that identifies a subset of the distribution of samples based on the uncertainty score, and an explanation component that generates an explanation of an output of the analytical model, corresponding to the selected input, based on use of a sample from the subset of the distribution of samples.
1 . A system, comprising:
a memory that stores computer executable components; and
a processor that executes at least one of the computer executable components that
iteratively, until an uncertainty score for a distribution of samples satisfies a defined confidence threshold and the distribution of samples fall on a straight line extending from a selected input:
selects, based on a defined selection process, a set of inputs from a set of samples neighboring the selected input, wherein the set of samples neighboring the selected input are scattered in a non-linear region surrounding the selected input,
generates, using an analytical model, a set of outputs based on the set of inputs,
determines the distribution of samples based on the set of outputs, and
determines the uncertainty score for the distribution of samples, wherein the uncertainty score indicates a level of confidence that relevant neighbor samples for the selected input are within the distribution of samples;
identifies a subset of the distribution of samples based on the uncertainty score; and
generates an explanation of an output of the analytical model generated using the selected input, based on use of a sample from the subset of the distribution of samples.
2 . The system of claim 1 , wherein the distribution of samples define local boundaries of a neighborhood surrounding the selected input.
3 . The system of claim 1 , wherein the determining the distribution of samples employs a multidimensional piecewise segmented regression.
4 . The system of claim 1 , wherein the at least one of the computer executable components further:
generates a proxy model based on fitting a regression model to the subset of the distribution of samples.
5 . The system of claim 4 , wherein the proxy model has a first algorithmic complexity that is lesser than a second algorithmic complexity of the analytical model.
6 . The system of claim 4 , wherein the at least one of the computer executable components further:
outputs from the proxy model a proxy output based on a sample input of at least one sample of the subset of the distribution of samples, and wherein the explanation is generated based on the proxy output.
7 . The system of claim 1 , wherein the system is analytical model agnostic.
8 . The system of claim 1 , wherein the at least one of the computer executable components further:
generates an image representing the explanation.
9 . A computer-implemented method, comprising:
iteratively, until an uncertainty score for a distribution of samples satisfies a defined confidence threshold and the distribution of samples fall on a straight line extending from a selected input:
selecting, by a system operatively coupled to a processor, based on a defined selection process, a set of inputs from a set of samples neighboring the selected input, wherein the set of samples neighboring the selected input are scattered in a non-linear region surrounding the selected input,
generating, by the system, using an analytical model, a set of outputs based on the set of inputs,
determining, by the system, the distribution of samples based on the set of outputs, and
determining, by the system, the uncertainty score for the distribution of samples, wherein the uncertainty score indicates a level of confidence that relevant neighbor samples for the selected input are within the distribution of samples;
identifying, by the system, a subset of the distribution of samples based on the uncertainty score; and
generating, by the system, an explanation of an output of the analytical model, generated using the selected input, based on use of a sample from the subset of the distribution of samples.
10 . The computer-implemented method of claim 9 , wherein the distribution of samples define local boundaries of a neighborhood surrounding the selected input.
11 . The computer-implemented method of claim 9 , wherein the determining the distribution of samples employs a multidimensional piecewise segmented regression.
12 . The computer-implemented method of claim 11 , wherein the multidimensional piecewise segmented regression comprises a multidimensional piecewise linear segmented regression or a multidimensional piecewise polynomial segmented regression.
13 . The computer-implemented method of claim 9 , further comprising:
generating, by the system, a proxy model based on fitting a regression model to the subset of the distribution of samples.
14 . The computer-implemented method of claim 13 , further comprising:
outputting, by the system, from the proxy model a proxy output based on a sample input of at least one sample of the subset of the distribution of samples, and wherein the explanation is generated based on the proxy output.
15 . A computer program product facilitating a process for providing an explanation result for an analytical model, the computer program product comprising a non-transitory computer readable medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
iteratively, until an uncertainty score for a distribution of samples satisfies a defined confidence threshold and the distribution of samples fall on a straight line extending from a selected input:
select, based on a defined selection process, a set of inputs from a set of samples neighboring the selected input, wherein the set of samples neighboring the selected input are scattered in a non-linear region surrounding the selected input,
generate, using the analytical model, a set of outputs based on the set of inputs,
determine the distribution of samples based on the set of outputs, and
determine the uncertainty score for the distribution of samples, wherein the uncertainty score indicates a level of confidence that relevant neighbor samples for the selected input are within the distribution of samples;
identify a subset of the distribution of samples based on the uncertainty score; and
generate an explanation of an output of the analytical model generated using the selected input, based on use of a sample from the subset of the distribution of samples.
16 . The computer program product of claim 15 , wherein the distribution of samples define local boundaries of a neighborhood surrounding the selected input.
17 . The computer program product of claim 16 , wherein the determining the distribution of samples employs a multidimensional piecewise segmented regression.
18 . The computer program product of claim 17 , wherein the multidimensional piecewise segmented regression comprises a multidimensional piecewise linear segmented regression or a multidimensional piecewise polynomial segmented regression.
19 . The computer program product of claim 15 , wherein the program instructions are further executable by the processor to cause the processor to:
generate, by the processor, a proxy model based on fitting a regression model to the subset of the distribution of samples.
20 . The computer program product of claim 19 , wherein the program instructions are further executable by the processor to cause the processor to:
output, by the processor, from the proxy model a proxy output based on a sample input of at least one sample of the subset of the distribution of samples, and wherein the explanation is generated based on the proxy output.