Loyalty Extraction Using QMS and Eclectic Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning classifiers often suffer from extreme bias due to focusing on specific aspects, neglecting others, and struggle with complex classification problems where a single inference function fails to accurately distinguish characteristics among elements.
Innovation Solution
The proposed method employs multiform separation (MS), specifically quadratic multiform separation (QMS), combined with an eclectic classifier, using multiple piecewise continuous functions and weighted cost functions to generate multiple classification results, which are then analyzed by an eclectic classifier to achieve a final decision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single classifier model is used, then the model structure is simple, but the classification accuracy and robustness deteriorate due to extreme bias
Solution Approach 1:
The patent divides the classification task into multiple independent sub-tasks by using multiple classifier models (e.g., CNN for image classification, SVM for pattern recognition). Each model focuses on specific characteristics of the data, and their results are combined through a meta-classifier to produce the final classification decision. This segmentation allows each model to specialize in particular aspects without being constrained by a single model's limitations.
Solution Approach 2:
The patent merges multiple classifier models and their outputs into a unified classification system. The meta-classifier integrates predictions from multiple base classifiers, weighing their contributions based on performance metrics. This combining approach leverages the strengths of different models while compensating for their individual weaknesses, thereby improving overall classification accuracy and robustness.
2Device complexity
If a single inference function is used, then the system is simple, but it cannot accurately distinguish characteristics among elements in complex classification problems
Solution Approach 1:
The patent segments the classification problem into multiple sub-problems, each handled by a specialized inference function. For example, one function might detect spatial patterns while another detects temporal patterns. Each inference function is optimized for specific characteristics, enabling precise distinction that a single function cannot achieve alone.
Solution Approach 2:
The patent introduces an additional dimension by using multiple inference functions that operate in different feature spaces or transformation domains. Instead of relying on a single function to capture all characteristics, the system transforms data through multiple functions, each revealing different aspects, and then integrates these multi-dimensional perspectives for comprehensive classification.
3Reliability
If multiple classifier models are used, then the classification accuracy and robustness improve, but the system complexity and computational cost increase
Solution Approach 1:
The patent designs a universal meta-classifier framework that can accommodate multiple base classifier models with different architectures and training methods. The meta-classifier serves as a universal interface that standardizes the integration of diverse models, managing system complexity through abstraction. This multi-functionality allows the system to leverage various models without proportionally increasing operational complexity.
Solution Approach 2:
The patent implements feedback mechanisms where the performance of individual classifier models is continuously evaluated and used to adjust their weights and contributions in the meta-classifier. Models that perform well on specific datasets or characteristics receive higher weights, while underperforming models are downweighted or replaced. This feedback loop optimizes the system dynamically, improving accuracy while managing complexity through adaptive model selection.
Data Source
AI summary
The present invention provides a loyalty extraction machine, wherein “quadratic multiform separation” (QMS) is modified and executed multiplicatively in an even generalized way. In each execution, the characteristic of one single membership is either enhanced or reduced. This process is performed in turn to each membership. Thus, every sample data (or element) receives multiple classification results. Then, the multiple classification results are collected and analyzed by an “eclectic classifier” to reach a final decision. The combination of the generalized QMS and the eclectic classifier therefore develops the loyalty extraction machine. Moreover, a label called “loyalty type” of the element is introduced to describe the effectiveness of membership recognition with respect to a training set.