Data Classifier Using Proximity Graphs and Weighted Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data classification methods are inefficient in handling large datasets, requiring excessive time and computational resources, and often yield less-than-desired classification accuracy due to suboptimal query selection criteria.
Innovation Solution
A data classifier device and method that uses a passive classifier and query engine to predict classification labels by determining relationships between data and training data, assigning weighted labels based on similarity in a proximity graph, and actively selecting additional data points to augment the training set, thereby reducing processing time and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing active classifier algorithms are used to predict class labels for large datasets, then classification accuracy can be achieved, but the time and computational resources required become excessively large
Solution Approach 1:
The patent segments the classification task into two distinct components: a passive classifier that performs rapid initial classification of all data points, and an active classifier that selectively refines only the most uncertain cases. This segmentation allows the system to achieve high accuracy through the active component while maintaining fast processing through the passive component handling the bulk of classifications.
Solution Approach 2:
The patent applies partial action by having the active classifier focus only on a subset of data points that require refinement, rather than processing the entire dataset. The passive classifier handles the majority of classifications efficiently, and the active classifier applies its computationally intensive algorithms only where needed, significantly reducing overall processing time while maintaining accuracy.
2Quantity of substance
If existing active classifier algorithms are used for classification, then labels can be predicted with limited training data, but the computational resources required become excessively large
Solution Approach 1:
The patent segments the computational workload between a passive classifier that uses simple, resource-efficient methods for initial classification, and an active classifier that uses more resource-intensive algorithms only on uncertain cases. This segmentation reduces overall computational resource consumption while maintaining the ability to work with limited training data.
Solution Approach 2:
The passive classifier serves as an intermediary that preprocesses data and identifies uncertain cases before they reach the active classifier. This intermediary step filters out most data points that don't require intensive processing, reducing the computational burden on the active classifier while preserving the ability to achieve accurate classifications with limited training data.
3Adaptability or versatility
If existing active classifier algorithms use different criteria for classification and query selection, then the system can operate with flexible criteria, but classification accuracy becomes less than desired
Solution Approach 1:
The patent merges the classification function and query selection function into a unified framework where both the passive and active classifiers use consistent criteria for determining class labels. The passive classifier's simple criteria and the active classifier's refined criteria work together coherently, eliminating the accuracy degradation that occurs when different criteria are used separately.
Solution Approach 2:
The patent implements feedback mechanisms where the active classifier's results inform and refine the passive classifier's criteria over time. This feedback loop ensures that both classifiers progressively align their criteria, improving classification accuracy while maintaining the flexibility to adapt to different data distributions and classification tasks.
Data Source
AI summary
An illustrative data classifier device includes data storage and at least one processor configured to operate as a query engine and a passive classifier that is configured to predict classification labels for data. The processor is configured to determine a relationship between the data and training data with associated training classification labels. The processor is also configured to assign a weighted version of at least one of the training classification labels to at least one member of the data based on the determined relationship. An illustrative method of classifying data includes predicting classification labels for data by determining a relationship between the data and training data with associated training classification labels. A weighted version of at least one of the training classification labels is assigned to at least one member of the data based on the determined relationship.