Query Object Resemblance Evaluation via Disjoint Region Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing decision aid systems for object classification lack intelligibility, as users are required to understand complex methods and processes, leading to a lack of confidence in decisions made by these systems, and visualization methods often result in false neighborhoods that distort the similarity measurements between objects.
Innovation Solution
A method that projects reference data into a lower-dimensional space, partitions it into disjoint regions, and evaluates the class of a test datum based on similarity measurements, allowing users to visualize the degree of resemblance between the query object and reference objects without positioning the query object on the map, thereby providing a stable mental representation of the reference object universe.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If decision aid systems provide confidence indicators through complex probabilistic methods or multiple discriminators, then the quality of decision support is improved, but the intelligibility to non-specialist users deteriorates
Solution Approach 1:
The patent extracts the query object from the projection space and evaluates its class independently by comparing similarity measurements to reference objects, rather than requiring users to interpret complex projection positions or confidence indicators. This separates the decision support function from the complex visualization methodology.
Solution Approach 2:
The patent changes the parameter representation from abstract projection coordinates and confidence scores to concrete similarity measurements between the query object and reference objects. This transformation makes the decision criteria directly intelligible to users while maintaining decision quality.
2Ease of operation
If visualization methods project objects onto a map to show similarity, then spatial organization is improved, but false neighborhoods are created that distort similarity measurements
Solution Approach 1:
The patent extracts the query object from the projection space entirely, avoiding the false neighborhood problem by not positioning the query object on the map. Instead, it evaluates class membership by comparing similarity measurements to reference objects whose positions are already established.
Solution Approach 2:
Rather than positioning the query object on the map and measuring its distance to reference objects, the patent inverts the approach by keeping reference objects fixed on the map and evaluating the query object's class through similarity comparisons without spatial positioning.
3Adaptability or versatility
If discriminators assign class membership probabilities to multiple classes, then classification flexibility is improved, but the decision-making process becomes less transparent to users
Solution Approach 1:
The patent extracts the similarity measurement information directly for evaluation, presenting it in a transparent format that shows users exactly how the classification decision is derived from comparisons with reference objects, rather than hiding it in probability calculations.
Solution Approach 2:
The patent uses similarity measurements as an intermediary between the query object and class membership determination. This intermediary provides a transparent, interpretable metric that users can understand while still enabling flexible classification across multiple classes.
Data Source
AI summary
A method and a system for evaluating the class of a test datum in a data space of dimension D where D≧3, each datum belonging to at least one class grouping together several data, comprising: projecting a suite of reference data of the data space into a space of dimension Q where Q<D, the class of each reference datum being known, calculating a measurement of similarity of the test datum to each of the reference data, partitioning the projection space into a plurality of disjoint regions each containing the projection of one and only one reference datum, and finally evaluating the class of the test datum, this class being evaluated as being the same class as one of the reference data contained in one of the regions containing the reference data closest to the test datum in the sense of the similarity measurement.