Diffusion Model Error Mitigation in Circuit-Cut Quantum Circuits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Quantum computing is limited by noise accumulation during circuit cutting and knitting, which renders final results unusable, and traditional Quantum Error Mitigation (QEM) methods incur high computational costs due to additional circuit executions.
Innovation Solution
Integrate machine learning-based Quantum Error Mitigation (ML-QEM) with circuit cutting and knitting processes to predict noise-free outputs without additional circuit executions, using a trained ML model to estimate noise-free outputs for each subcircuit and combine them efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional Quantum Error Mitigation (QEM) is used to reduce noise, then noise is mitigated, but additional circuit executions are required which increase computational cost
Solution Approach 1:
The patent trains an ML model in advance on noise-free quantum circuit data to learn the mapping between noisy and noise-free outputs. This preliminary training phase allows the model to predict noise-free results during actual quantum computations without requiring additional circuit executions, thus resolving the contradiction between noise mitigation and computational efficiency
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between noisy quantum circuit outputs and the desired noise-free results. The model acts as a mediator that translates noisy measurements into accurate predictions, eliminating the need for traditional QEM methods that require additional circuit executions
2Device complexity
If circuit cutting is used to divide large quantum circuits into subcircuits, then quantum computer limitations are overcome, but noise accumulates during knitting operations making final results unusable
Solution Approach 1:
The patent applies the ML model to each subcircuit individually before the knitting operation. By predicting noise-free outputs for each subcircuit in advance, the accumulated noise from multiple subcircuit executions and knitting operations is eliminated, allowing large quantum circuits to be divided and recombined without losing result accuracy
Solution Approach 2:
The patent replaces the traditional mechanical knitting operation that physically combines subcircuit results (and accumulates noise) with an ML-based prediction system. The ML model substitutes the noisy knitting process by directly predicting what the final result would be, thus resolving the contradiction between handling large circuits and maintaining result accuracy
Data Source
AI summary
One example method includes performing, on a quantum circuit, a circuit cutting process to generate a set of subcircuits of the quantum circuit, for each of the subcircuits, (1) running the subcircuit on a target backend quantum hardware configuration, and measuring a noisy output resulting from the running of the subcircuit, (2) using an ML (machine learning) model to obtain an estimate of a noise-free output of the subcircuit, and (3) replacing noisy output of the subcircuits with the estimated noise-free output generated by the ML model. The method further includes running a circuit knitting process, using the estimates, to knit the subcircuits together to form the quantum circuit, and executing the quantum circuit that resulted from the circuit knitting process.