Directional Nearest Neighbor Detection Using Randomized Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional nearest neighbor search techniques fail to effectively consider directional differences in feature distances, leading to inefficient calculation times and inability to distinguish between directions, especially in higher-dimensional spaces.
Innovation Solution
A similarity detecting apparatus and directional nearest neighbor detection method that utilize random number generation, key calculation functions, and table-based data management to prioritize searches along specific directions, reducing unnecessary calculations by registering data based on directional parameters and intensity weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional nearest neighbor search techniques are used, then data similarity can be detected, but calculation time increases significantly and directional differences are not considered
Solution Approach 1:
The patent divides the search space into multiple directional sectors by generating random directional parameters. Data is organized into tables based on these directional sectors, allowing the system to search only relevant directional regions rather than scanning all data uniformly, thus reducing calculation time while maintaining directional precision.
Solution Approach 2:
The patent applies different search strategies based on directional importance. By weighting directional parameters and creating direction-specific tables, the system focuses computational resources on directions that are more important for similarity detection, improving both speed and precision in those critical directions.
2Productivity
If all data is searched uniformly, then complete search coverage is achieved, but computational efficiency decreases in higher-dimensional spaces
Solution Approach 1:
The patent pre-processes data by generating random directional parameters and organizing data into tables based on these directions before actual search queries. This preliminary organization allows fast lookup during search operations without requiring complex real-time calculations, improving search speed while keeping the algorithm manageable.
Solution Approach 2:
The patent introduces random directional parameters as intermediary structures that mediate between the query data and the stored data. These directional parameters act as keys to access pre-organized data tables, simplifying the search process and improving efficiency in higher-dimensional spaces without requiring complex direct comparisons.
3Measurement precision
If directional parameters are introduced to improve search precision, then nearest neighbor detection accuracy improves, but system complexity increases
Solution Approach 1:
The patent changes the parameters used for data organization by introducing random directional parameters and intensity parameters. These parameter changes allow the system to reorganize data into direction-specific tables, improving nearest neighbor detection accuracy while the parameter-based approach keeps the complexity manageable through mathematical transformations rather than complex structural changes.
Data Source
AI summary
In order to detect similar data from a great deal of data at high speed, a similarity detecting apparatus includes a random number generating unit 3 which calculates random number data based on a parameter list, a table managing unit 5 which calculates a plurality of key calculation functions based on the random number data, calculates a plurality of tables based on input data, and calculates a candidate data list based on a query shown by a search condition, and a data processing unit 6 which calculates search result data from the candidacy data list to satisfy the condition shown a search condition. Each table is calculated for a value corresponding to a key to show a data list of data in which the data substituted into a key calculation function is equal to the key. A candidate data list contains a plurality of search data list and a search data list corresponding to the table shows the value corresponding to a query value obtained by substituting the query into the key calculation function.