Ising Machine Coefficient Memory Retention for Optimization Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information processing systems that utilize Ising machines for combinatorial optimization problems face inefficiencies due to the large volume of data in coefficient matrices and vectors, leading to prolonged data transfer times and reduced overall throughput.
Innovation Solution
The system employs an Ising machine and a host apparatus that execute multiple search processes with varying initial values for the Ising spins, allowing for the calculation of different approximate solutions and selecting the best one, while minimizing the need for frequent data transfers by storing previous coefficient matrices and vectors until new ones are written, thus improving data transfer efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the coefficient matrix and coefficient vector are transmitted from the host apparatus to the Ising machine for each search process, then the Ising machine can perform optimization searches, but the data transfer time increases and system throughput decreases
Solution Approach 1:
The coefficient matrix and coefficient vector are transmitted to the Ising machine before multiple search processes are executed. The host apparatus sends the problem definition data once, and the Ising machine retains it in its memory for subsequent search operations, eliminating the need for repeated data transfers.
Solution Approach 2:
The Ising machine's coefficient memory is designed to store and reuse the coefficient matrix and coefficient vector across multiple search processes. This allows the same problem definition to be used for multiple optimization searches with different initial values, making the data transmission infrastructure serve multiple functions.
2Measurement precision
If multiple search processes are executed with different initial values to find better solutions, then solution quality improves, but the frequency of data transfers increases and throughput decreases
Solution Approach 1:
The problem definition data (coefficient matrix and coefficient vector) is prepared and transmitted in advance before multiple search processes are initiated. This allows the system to execute multiple searches with different initial values without repeating the data transmission step, thereby maintaining high throughput while improving solution quality.
Solution Approach 2:
The Ising machine continuously performs multiple search processes using the same coefficient data without interruption for re-transmission. The machine executes successive optimization searches with varying initial values, maintaining continuous productive operation while seeking better solutions.
Data Source
AI summary
An information processing system according to one embodiment includes an Ising machine and a host apparatus. The Ising machine includes a coefficient memory, a variable memory, an arithmetic circuit, an output circuit, and a setting circuit. The coefficient memory stores a coefficient matrix and a coefficient vector defining an Ising model. The variable memory stores a main variable and an auxiliary variable corresponding to each of Ising spins contained in the Ising model. The arithmetic circuit alternately repeats, in a search process, execution of an auxiliary variable update process updating the auxiliary variable with the main variable and a main variable update process updating the main variable with the auxiliary variable, for each of the Ising spins. The coefficient memory continues storing a preceding coefficient matrix and a preceding coefficient vector until the setting circuit writes a new coefficient matrix and a new coefficient vector.