Heterogeneous Decision Engine Label Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional multiple classifier systems (MCS) are limited by requiring homogeneous labels from decision engines, making it difficult to integrate legacy decision engines with non-uniform labels, which are often used in different contexts and have varying granularities, thus preventing effective decision fusion.
Innovation Solution
The technique involves normalizing heterogeneous decision engine outputs through knowledge-based label unification, user selection, and data-based value calculation to generate unified class labels and corresponding values, allowing for decision fusion across engines with different label sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional MCS requires homogeneous labels from decision engines, then decision fusion can be performed, but legacy decision engines with heterogeneous labels cannot be integrated
Solution Approach 1:
The patent introduces an intermediary label normalization module that mediates between heterogeneous decision engines and the decision fusion mechanism. This module receives diverse labels from multiple decision engines, standardizes them into a unified format, and then feeds them to the decision fusion process, enabling legacy engines to be integrated without requiring modifications to their label schemas
Solution Approach 2:
The patent transforms the label parameter from heterogeneous formats (different schemas, granularities, and terminologies) into a unified standardized format. By changing the parameter representation through mapping and normalization processes, the system enables compatibility across diverse decision engines while preserving their individual decision-making capabilities
2Adaptability or versatility
If multiple decision engines with different label granularities are combined, then more diverse opinions are available for decision fusion, but label unification becomes difficult
Solution Approach 1:
The patent resolves label granularity conflicts by introducing an additional dimension of abstraction. Instead of forcing labels to match at the same granularity level, the system creates a hierarchical mapping structure that accommodates different granularities (fine-grained, medium-grained, coarse-grained labels) and maps them to a unified decision space, preserving the diversity of opinions while enabling fusion
3Reliability
If legacy decision engines are integrated into MCS, then system accuracy improves through combined opinions, but the system complexity increases due to heterogeneous label handling
Solution Approach 1:
The patent extracts the label normalization function as a separate, dedicated module within the MCS architecture. By isolating the complexity of heterogeneous label handling into this specialized component, the rest of the system (decision engines and fusion mechanism) can operate with simplified, standardized labels, thereby improving reliability through diverse engine integration while containing system complexity in a manageable location
Data Source
AI summary
A method for using a plurality of decision engines to produce a single decision comprises: receiving heterogeneous data from the plurality of decision engines comprising a set of class labels from each of the plurality of decision engines, wherein the set of class labels from at least a first decision engine differs from the set of class labels from at least a second decision engine; generating a single set of unified class labels from the heterogeneous data; calculating at least one value corresponding to each of at least a subset of the unified class labels; and performing decision fusion on at least the subset of the unified class labels and corresponding values to produce the single decision.