Decision Boundary Manager for Machine Learning Model Adaptation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Machine learning models often require significant time and resources to update decision boundaries in response to changing business rules, regulations, or biased training data, leading to outdated classifications.

Innovation Solution

A decision boundary manager system that receives feedback rules, generates an updated feedback rule set, and creates a new training dataset to retrain the model, allowing for the efficient movement of decision boundaries without the need for extensive new data accumulation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional machine learning models are retrained to update decision boundaries, then the model can adapt to new rules and regulations, but it requires significant time and resources

Engineering Contradiction:

Improveadaptability to new rulesVSAvoidtime to update decision boundaries

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by accumulating feedback rules and generating updated training datasets in advance, before full retraining is needed. The feedback rule set is built incrementally from incoming feedback, and updated training datasets are prepared beforehand, so when decision boundary updates are required, the model can be retrained quickly with pre-prepared data rather than collecting data from scratch.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system introduces an intermediary layer between the original training data and the model retraining process. This intermediary generates synthetic updated training datasets by applying feedback rules to existing training data, transforming the original dataset into an updated version that reflects new business rules without requiring external data collection or full model retraining.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If traditional machine learning models are retrained to update decision boundaries, then the model can adapt to new rules and regulations, but it requires significant resources

Engineering Contradiction:

Improveadaptability to new rulesVSAvoidresources to update decision boundaries

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates copies of the original training dataset and applies feedback rules to generate synthetic updated training datasets. Instead of requiring new real-world data collection, the system copies existing training data and transforms it according to feedback rules, significantly reducing the resources needed for model updates while maintaining adaptability to new business rules.

Inventive Principle:

Principle #26Copying

3Productivity

If feedback rules are applied directly without resolving conflicts, then the update process is faster, but the model may learn inconsistent patterns

Engineering Contradiction:

Improvespeed of updating decision boundariesVSAvoidconsistency of learned patterns

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system implements a feedback mechanism where conflicting feedback rules are detected and resolved through conflict analysis. When multiple feedback rules apply to the same training instance and produce conflicting outcomes, the conflict resolution logic identifies and resolves these conflicts before generating the updated training dataset, ensuring consistent learned patterns while maintaining update speed through automated conflict detection and resolution.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12106193B2Moving decision boundaries in machine learning models

Publication Date: 2024.10.01 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Embodiments are disclosed for a method. The method includes receiving feedback decision rules for multiple predictions by a trained machine learning model. generating a feedback rule set based on the feedback decision rules. The method further includes generating an updated training dataset based on an original training dataset and an updated feedback rule set. The updated feedback rule set resolves one or more conflicts of the feedback rule set, and the updated training dataset is configured to train the machine learning model to move a decision boundary. Generating the updated training dataset includes generating multiple updated training instances by applying one of the feedback decision rules to a training instance of the original training dataset.