Graph Neural Network for Quantum Circuit Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current quantum experimentation methods are cumbersome, time-consuming, and prone to errors due to manual efforts, lack of automation in identifying and comparing experiment combinations, and difficulty in tracking results, which hampers the efficient selection of appropriate quantum circuits.

Innovation Solution

A processor-implemented method and system that generates and prioritizes high-level combinations of quantum tasks using a graph-based approach, computing dissimilarity scores, soft-clustering vertices, and iteratively selecting experiments based on probability scores to optimize execution and train a Graph Neural Network (GNN) model for recommending optimal quantum circuit combinations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If manual experimentation approaches are used to identify and test quantum circuit combinations, then flexibility and adaptability are maintained, but the process becomes time-consuming, error-prone, and difficult to scale

Engineering Contradiction:
Improveautomation of quantum circuit experimentationVSAvoidcomplexity of experimentation system
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary system comprising a workflow engine, experiment manager, and result analyzer that mediates between the user's quantum circuit definitions and the actual experimentation process. This intermediary layer automates task generation, experiment execution, and result analysis, reducing manual intervention while managing system complexity through modular architecture.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces manual mechanical experimentation processes with automated computational systems. The workflow engine automatically generates quantum circuit configurations, the experiment manager executes them on quantum hardware or simulators, and the result analyzer processes outcomes—substituting manual trial-and-error with systematic automated exploration.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If all possible quantum circuit combinations are tested exhaustively, then complete exploration of the search space is achieved, but the time and computational resources required become prohibitive

Engineering Contradiction:
Improvecompleteness of experiment explorationVSAvoidtime for experimentation
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent implements partial action by using the Graph Neural Network to prioritize and select a subset of most promising circuit combinations for experimentation. Instead of exhaustively testing all possible configurations, the system identifies and executes only those experiments with highest predicted performance, achieving reliable results with fraction of the time required for complete exploration.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent applies preliminary action through the Graph Neural Network model that pre-evaluates and ranks quantum circuit combinations before actual experimentation. The GNN learns from historical experiment data to predict which circuits are most likely to succeed, allowing the system to pre-select promising candidates and avoid wasting time on unlikely configurations.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If multiple quantum circuit experiments are executed in parallel, then productivity is improved, but tracking and comparing results becomes more difficult

Engineering Contradiction:
Improvespeed of quantum circuit optimizationVSAvoidtracking and comparison of experiment results
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent implements a universal result analysis framework that handles multiple experiment outcomes through a single integrated system. The result analyzer component processes results from parallel experiments uniformly, normalizing different metric types (fidelity, success probability, execution time) into a common evaluation scale, enabling straightforward comparison across all parallel experiments regardless of their specific configurations.

Inventive Principle:
Principle #6Universality (Multi-functionality)

4Measurement precision

If Graph Neural Network is trained on comprehensive training data from quantum experiments, then recommendation accuracy is improved, but the training process becomes more computationally intensive

Engineering Contradiction:
Improveaccuracy of circuit combination recommendationsVSAvoidcomputational resources for GNN training
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies partial action in GNN training by using a curricula learning approach where the model is first trained on a subset of representative experiment data, then progressively refined with additional data. This staged training approach achieves high recommendation accuracy without requiring all possible training data to be processed simultaneously, reducing computational resource requirements while maintaining model performance.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4307173A1Method and system for recommending optimum combination of quantum circuits
Publication Date: 2024.01.17 TATA CONSULTANCY SERVICES LTD
  • EP4307173A1 patent drawingFigure 1
  • EP4307173A1 patent drawingFigure 2
  • EP4307173A1 patent drawingFigure 3A

AI summary

Traditional approaches for recommending optimum combination of quantum circuits are experimentation based approaches, and require manual efforts or are cumbersome, effort intensive and iterative processes. Method and system disclosed herein generally relates to quantum experimentation, and, more particularly, for recommending optimum combination of quantum circuits. In this approach, a high-level combination of experiments are initially generated, which are further prioritized using a graph based approach, which then forms a training data. The training data is then used for generating a GNN data model, which is further used for recommending optimum combination of quantum circuits.