Model Explanation Stress Testing for Change-Agnostic Data Points

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing prediction models are susceptible to changes under input data manipulation or small perturbations, leading to unstable explanations that may not meet regulatory standards, necessitating the development of change-agnostic data points that remain robust to such variations.

Innovation Solution

A data points computing module is implemented using processors and memory to compute change-agnostic data points by training machine learning models, applying sampling algorithms, and performing data perturbation tests to ensure model explanations remain stable under data changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If prediction models are used to analyze input data, then predictive accuracy is improved, but explanation stability deteriorates under small data perturbations

Engineering Contradiction:

Improvepredictive accuracyVSAvoidexplanation stability

Core Design Contradiction:

Measurement precisionVSStability of the object's composition

Solution Approach 1:

The system performs preliminary action by computing change-agnostic data points before actual prediction occurs. It generates alternative data points that are resistant to perturbations and uses these pre-computed stable data points to train the prediction model, ensuring that the model learns from stable representations rather than unstable original data.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system introduces change-agnostic data points as an intermediary between the original input data and the prediction model. These intermediary data points serve as a bridge that filters out noise and perturbations, allowing the model to receive cleaned, stable representations of the data without directly processing the unstable original data.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If complex machine learning models are used to improve prediction performance, then predictive capability is enhanced, but explanation robustness deteriorates under data manipulation

Engineering Contradiction:

Improvepredictive capabilityVSAvoidexplanation robustness

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system performs preliminary computation of change-agnostic data points before model training. By pre-computing stable representations of the data, the system ensures that even complex models will receive stable input for learning, thereby maintaining explanation robustness while allowing the use of complex predictive models.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates copies of the original data points that are specifically designed to be resistant to changes. These copied data points (change-agnostic data points) preserve the essential information needed for prediction while eliminating the vulnerability to perturbations that affects the original data.

Inventive Principle:

Principle #26Copying

3Measurement precision

If data points are used for model training, then model accuracy is improved, but sensitivity to perturbations increases

Engineering Contradiction:

Improvemodel accuracyVSAvoidsensitivity to perturbations

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The system converts the harmful effect of data perturbations into a beneficial process by using the perturbation resistance requirement to guide the selection and creation of change-agnostic data points. Instead of treating perturbations as noise to be ignored, the system uses this constraint to actively select and generate data points that are inherently resistant to such harmful variations.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Solution Approach 2:

The change-agnostic data points serve as an intermediary layer between the raw input data and the model training process. This intermediary transforms the data in a way that preserves accuracy information while filtering out perturbation sensitivity, allowing the model to learn from accurate representations without inheriting the harmful sensitivity to data variations.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12530623B2System, method, and computer program for computing change-agnostic data points through variability explanation stress testing

Publication Date: 2026.01.20 JPMORGAN CHASE BANK NA

AI summary

Various methods, apparatuses/systems, and media for computing change-agnostic data points are disclosed. A processor trains a machine learning model by using the at least the first set of raw data; computes a set of explanations for all combinations based on output data of the trained machine learning model, the first set of raw data, and sampled raw data computed by applying a sampling algorithm on the raw data; computes a compact representation of the set of explanations corresponding to a pre-configured dimension based on compression quality and generating a set of compressed explanations; computes a unique representation of model explanation with respect to the pre-configured dimension; determines whether the model explanation is robust to changes in data through data perturbation; and computes change-agnostic data points based on determining that the model explanation is robust to changes in data through data perturbation.