Class Prediction Using Fuzzy Feature Encoding and Entry Coalescing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As the multiplicity of features and possible classifications increases, the number of entries required in a lookup table for data classification becomes impractically large, especially in hardware implementations with limited storage capacity, such as portable devices.
Innovation Solution
The approach utilizes fuzzy logic to map feature data into encodings, forming joint feature items by concatenating these encodings, and uses class counters to determine predicted classes based on membership values, reducing storage requirements through entry coalescing and distance-based encodings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the multiplicity of features and possible classifications increases, then the classification capability and versatility improve, but the number of entries required in the lookup table increases making storage impractical
Solution Approach 1:
The patent segments the feature space by dividing each feature's range into multiple discrete regions. Instead of creating a lookup table entry for every possible feature combination, each feature is independently segmented into K regions, and only the region indices are stored and processed. This reduces the storage requirement from exponential (all feature combinations) to linear (one index per feature).
Solution Approach 2:
The patent transforms the problem from storing high-dimensional feature combinations into storing low-dimensional region indices. By mapping continuous feature values to discrete region indices, the system effectively changes the dimensionality of the stored data, enabling efficient storage and processing while maintaining classification capability.
2Measurement precision
If the number of features and classes increases to improve prediction accuracy, then the classification precision improves, but the device complexity and storage requirements become impractical for hardware implementations
Solution Approach 1:
The patent applies local quality by creating region-specific encodings for different feature ranges. Each feature is divided into regions with potentially different encoding schemes, allowing the system to adapt the representation to local characteristics of the data distribution while maintaining overall system efficiency.
Solution Approach 2:
The patent changes the parameter representation from continuous feature values to discrete region indices. This parameter transformation enables the system to maintain high prediction accuracy through sufficient region granularity while dramatically reducing storage requirements by working with compact integer indices instead of continuous values.
Data Source
AI summary
Apparatuses and methods for supporting class prediction based on multiple items of feature data are provided. Learning phase training data with known classification are used as inputs. Each event of the training data maps multiple items of feature data to encodings, where a range of values for each feature input are mapped to a given encoding. The concatenated encoding for the event form a joint feature item. Class counters are used to count class known to associated with the training event for the joint feature item in a table. At the conclusion of the training phase the class counter values enable a predicted class to be associated with each joint feature item in the table. In the inference phase the table is used for class prediction generation for new data events. The inference phase may be implemented in hardware which has less data handling capability than in the learning phase.