Matrix operation program, matrix operation method, and information processing device.
By dividing matrices into cache-compatible blocks and utilizing multiple CPU cores for parallel processing, the computational complexity of pattern mining is reduced, enhancing the speed of large-scale data analysis.
JP7893103B2Active Publication Date: 2026-07-22FUJITSU LTD
View PDF 7 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- FUJITSU LTD
- Filing Date
- 2022-09-15
- Publication Date
- 2026-07-22
AI Technical Summary
Technical Problem
The computational complexity of pattern mining increases exponentially as the number of products being combined increases, leading to significant processing challenges in large-scale datasets.
Method used
The matrix operation program divides the matrix into blocks of columns based on the storage capacity of the L2 cache, allowing parallel processing of these blocks using multiple CPU cores to reduce cache misses and accelerate pattern mining.
Benefits of technology
This approach significantly reduces memory access time and accelerates pattern mining processes in large-scale datasets by minimizing cache misses and optimizing storage utilization.
✦ Generated by Eureka AI based on patent content.
Smart Images

Figure 0007893103000001 
Figure 0007893103000002 
Figure 0007893103000003
Abstract
To increase a speed of a patterning mining process.SOLUTION: A matrix operation program according to an embodiment causes a computer to execute a divide process and an execute process. In a matrix operation in which a process of combining at least two columns included in a matrix and finding a product of each row of the combined columns to calculate a sum total of the products of each row is executed on all combinations of the columns in the matrix by an arithmetic unit, the divide process divides the matrix into blocks of column groups on the basis of the data size of the columns and the storage capacity of a second storage unit that the arithmetic unit accesses before a first storage unit that stores information regarding the matrix so that the column groups to be combined fit in the second storage unit. The execute process executes a process of calculating for each block of divided column groups.SELECTED DRAWING: Figure 10
Need to check novelty before this filing date? Find Prior Art