Database Circuit Matching With Adaptive Reference Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D memory systems face challenges in improving the accuracy of in-memory computing while reducing computational complexity and time, particularly in data searching and matching processes.
Innovation Solution
A database circuit comprising multiple data circuits and a processing circuit that stores and matches feature data through various configurations, using a reference value system to determine matching results based on similarity values and bit differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional data searching methods are used in 3D memory systems, then the search process is simple to implement, but the computational time and complexity increase significantly
Solution Approach 1:
The patent replaces traditional mechanical/CPU-based data searching with in-memory computing that performs search operations directly within the 3D memory structure. The memory cells themselves participate in the computation by generating matching signals based on query data comparison, eliminating the need for external computational processing and significantly reducing search time and complexity.
Solution Approach 2:
The 3D memory structure is designed to perform multiple functions: it serves as both data storage and data processing unit. The memory cells can store feature data and simultaneously perform matching operations with query data, generating matching signals that indicate similarity. This multi-functionality eliminates the need for separate processing units and reduces overall system complexity.
2Productivity
If in-memory computing is used to reduce computational time, then the searching speed improves, but the accuracy of data matching decreases
Solution Approach 1:
The patent employs multiple reference values (first reference value, second reference value) to dynamically adjust the matching threshold based on the distribution of matching signals. This parameter adjustment mechanism allows the system to maintain high accuracy by adapting the decision threshold to the specific query and data set, while still operating at high speed through in-memory computing.
Solution Approach 2:
The system uses feedback from the distribution of matching signals to determine the appropriate reference value for making match decisions. By analyzing the collected matching signals and comparing them against reference values, the system can adjust its decision-making process to maintain high accuracy. The feedback mechanism ensures that the in-memory computing operations produce reliable and accurate matching results.
Data Source
AI summary
A database circuit comprising data circuits and a processing circuit is provided. The data circuits are configured to receive a query data and respectively store a plurality of feature data that form at least one feature data class. The processing circuit is coupled to the data circuits, and configured to match the query data with the feature data to generate matching signals and take a first reference value as a target reference value to compare the matching signals with the target reference value. When at least one of the matching signals is greater than or equal to the target reference value, the processing circuit selects one of the at least one feature data class as a matching result; otherwise, the processing circuit takes a second reference value lower than the first reference value as the target reference value to compare the matching signals with the target reference value.