Feature Dataset Classification via Lookup Table Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems, such as those using Gaussian naïve Bayes algorithms, require significant computational resources to classify input data sets, making them unsuitable for devices with limited data processing capabilities.
Innovation Solution
The approach involves generating multiple distinct classifiers for each feature and selecting the representative class based on precomputed class probabilities, reducing the computational load by using class indications stored in lookup tables and enabling faster class determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Gaussian naïve Bayes algorithm is used for classification, then classification accuracy is improved, but computational resource requirement increases
Solution Approach 1:
The patent pre-computes class probabilities for each feature value permutation during an offline training phase and stores them in lookup tables. During online classification, the system simply retrieves pre-computed probabilities based on input feature values and performs lightweight aggregation, eliminating the need for complex real-time probability calculations while maintaining classification accuracy
Solution Approach 2:
The patent creates simplified copies of the classification model by storing pre-computed class probabilities in lookup tables rather than implementing the full Gaussian naïve Bayes calculation engine. This copy approach allows the system to achieve accurate classification by table lookup and simple aggregation, dramatically reducing computational requirements for devices with limited processing capability
2Measurement precision
If full precision representation is used for feature data values, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent changes the precision parameter of feature data representation to match the capabilities of low-complexity devices. By accepting that lower precision representations (fewer bits per feature value) are sufficient for the application, the system can use simpler hardware without lookup tables while still achieving accurate classification through the pre-computed probability approach
Data Source
AI summary
Apparatuses and methods of operating such apparatuses are disclosed. An apparatus comprises feature dataset input circuitry to receive a feature dataset comprising multiple feature data values indicative of a set of features, wherein each feature data value is represented by a set of bits. Class retrieval circuitry is responsive to reception of the feature dataset from the feature dataset input circuitry to retrieve from class indications storage a class indication for each feature data value received in the feature dataset, wherein class indications are predetermined and stored in the class indications storage for each permutation of the set of bits for each feature. Classification output circuitry is responsive to reception of class indications from the class retrieval circuitry to determine a classification in dependence on the class indications. A predicated class may thus be accurately generated from a simple apparatus.