Automated QUBO Generation from High-Level Source Code

Solution Overview

Problem

Existing methods for translating high-level source code into Quadratic Unconstrained Binary Optimization (QUBO) representations are inefficient, especially for higher-order functions, as they require manual expertise and are time-consuming due to the complexity of building coefficients, making them impractical for large-scale problems and non-expert users.

Innovation Solution

A computer-implemented method that automatically translates modular high-level source code into a QUBO representation by analyzing decision variables, applying encodings, generating polynomial portions, and reducing order terms, thereby creating a polynomial formulation suitable for input into information processing apparatuses like Ising machines.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual coefficient-by-coefficient construction is used for high-order QUBO functions, then correctness can be achieved, but the time consumption and complexity increase dramatically

Engineering Contradiction:

Improvecorrectness of QUBO formulationVSAvoidtime consumption for coefficient construction

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent segments the QUBO formulation process into independent subfunction translations. Each subfunction is translated separately using automated code analysis, and the results are composed together. This divides the complex O(N^k) coefficient construction into manageable pieces, maintaining correctness while reducing manual intervention time.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs self-service by automatically analyzing source code to extract decision variables, their types, and relationships. The automated variable analysis and encoding processes eliminate the need for manual coefficient construction, allowing the system to generate correct QUBO formulations without expert intervention.

Inventive Principle:

Principle #25Self-service

2Manufacturing precision

If manual expertise is required for QUBO creation from source code, then complex problems can be solved accurately, but the ease of operation decreases for non-expert users

Engineering Contradiction:

Improveaccuracy of optimization formulationVSAvoidusability for non-expert users

Core Design Contradiction:

Manufacturing precisionVSEase of operation

Solution Approach 1:

The system performs self-service by automatically analyzing source code to extract decision variables, their types, and relationships. The automated variable analysis and encoding processes eliminate the need for manual coefficient construction, allowing the system to generate correct QUBO formulations without expert intervention.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent creates a universal framework that works with any high-level source code containing optimization problems. The system handles multiple variable types (binary, continuous, integer) and different problem structures through a unified automated process, making it accessible to non-experts across diverse application domains.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If the number of coefficients is O(N^k) for k-th order functions, then the model can represent complex relationships, but the device complexity and processing requirements increase

Engineering Contradiction:

Improveability to represent complex relationshipsVSAvoidcomplexity of coefficient processing

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the QUBO formulation process into independent subfunction translations. Each subfunction is translated separately using automated code analysis, and the results are composed together. This divides the complex O(N^k) coefficient construction into manageable pieces, maintaining correctness while reducing manual intervention time.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces manual mechanical coefficient construction with automated computer-based analysis. The system uses source code analysis, variable encoding, and algorithmic composition to generate QUBO formulations, substituting automated computational processes for manual mathematical operations and reducing human cognitive load.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11481198B2Framework for automated generation of QUBO or higher order formulations from high-level source code

Publication Date: 2022.10.25 FUJITSU LTD

AI summary

Method of inputting code into information processing apparatus requiring code of polynomial format of order k which comprises translating the code into a polynomial of order k; encoding each decision variable in the code, and obtaining mappings between the encoded variables and the polynomial variables; extracting an expression representing interrelationships between the decision variables and polynomial subfunctions; generating first polynomial portions for subfunctions whose output is not the input for another subfunction. The method includes generating second polynomial portions for subfunctions whose output is the input for another subfunction, reducing the order of terms higher than floor(k/2) in the second polynomial portion to floor(k/2) with associated order reduction penalties, and creating an auxiliary variable equal to the second polynomial portion with an auxiliary penalty; and obtaining a polynomial formulation for input into the apparatus by accumulating the first polynomial portions, auxiliary variables and penalties, and order reduction penalties.