Information Processing System Encoding Coefficients for Combinatorial Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information processing systems face challenges in efficiently solving large-scale combinatorial optimization problems due to the exponential increase in solution states, leading to long data transfer times and reduced accuracy when using high-accuracy, wide-value-range Ising models, and similar issues with other optimization problems like QUBO and HUBO.
Innovation Solution
An information processing system that encodes coefficients into smaller identification values, allowing for faster data transfer while maintaining accuracy by using restoration information to restore the original coefficient values during the solving process, thereby reducing data size and storage requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the Ising model is expressed by a data type with high accuracy and a wide value range, then the versatility of the Ising model is improved, but the data transfer time from the host unit to the Ising machine increases
Solution Approach 1:
The coefficient data is segmented into two parts: identification values (stored in the Ising machine's memory) and restoration values (stored in the host unit's memory). This segmentation allows the Ising machine to work with compact identification values during computation, reducing data transfer time, while the host unit retains the full-precision restoration values to maintain accuracy and versatility when needed.
Solution Approach 2:
Identification values act as an intermediary between the host unit and the Ising machine. These identification values serve as compact proxies that represent the full-precision coefficients, enabling efficient data transfer and storage in the Ising machine while preserving the ability to reconstruct accurate coefficient values when required.
2Loss of time
If the Ising model is expressed by a data type with low accuracy and a narrow value range, then the data transfer time from the host unit to the Ising machine is shortened, but the versatility of the Ising model is deteriorated
Solution Approach 1:
The coefficient data is segmented into two parts: identification values (stored in the Ising machine's memory) and restoration values (stored in the host unit's memory). This segmentation allows the Ising machine to work with compact identification values during computation, reducing data transfer time, while the host unit retains the full-precision restoration values to maintain accuracy and versatility when needed.
Solution Approach 2:
The system dynamically changes the parameter representation: coefficients are transformed into identification values with smaller data sizes for storage and transfer in the Ising machine, while the ability to restore them to original high-precision parameters is preserved in the host unit. This parameter transformation enables the system to switch between efficiency mode (using identification values) and accuracy mode (using restoration values) as needed.
3Loss of time
If the data size of coefficients is reduced to shorten transfer time, then the data transfer time is shortened, but the accuracy of representing coefficients is reduced
Solution Approach 1:
The coefficient data is segmented into two parts: identification values (stored in the Ising machine's memory) and restoration values (stored in the host unit's memory). This segmentation allows the Ising machine to work with compact identification values during computation, reducing data transfer time, while the host unit retains the full-precision restoration values to maintain accuracy and versatility when needed.
Solution Approach 2:
The system dynamically changes the parameter representation: coefficients are transformed into identification values with smaller data sizes for storage and transfer in the Ising machine, while the ability to restore them to original high-precision parameters is preserved in the host unit. This parameter transformation enables the system to switch between efficiency mode (using identification values) and accuracy mode (using restoration values) as needed.
Data Source
AI summary
An information processing system according to an embodiment includes an information processing device and a solver device. The information processing device generates a plurality of coefficients based on a combinatorial optimization problem, encodes a first coefficient represented by a predetermined first data type into a first identification value represented by a second data type having a smaller data size than the first data type, outputs, to the solver device, coefficient information including the first identification value instead of the first coefficient, acquires ground state information from the solver device, and outputs a solution of the combinatorial optimization problem based on the ground state information. In the solving process, the solver device executes an operation on the first coefficient by using a first restoration value corresponding to the first identification value represented by the restoration information. The first restoration value is expressed by the first data type.