Dynamic Qubit Mapping for Quantum Circuit Transpilation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quantum computing transpilation processes often result in sub-optimal qubit mappings, leading to poor execution quality, lack of fine-grained control, and uncertainty about optimal mappings for accurate results, with errors accumulating due to fixed mappings during multiple circuit shots.
Innovation Solution
Implementing a method where qubit mappings are randomly reassigned after each circuit shot based on the relative quality of physical qubits, using a probability mass function to select mappings that mitigate accumulated errors and improve transpilation quality without excessive computational overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transpilation is performed once prior to circuit execution with fixed optimization objective, then transpilation computational overhead is reduced, but circuit execution quality deteriorates due to sub-optimal mappings
Solution Approach 1:
The system performs transpilation periodically at predetermined intervals during circuit execution rather than once at the beginning. This allows the quantum processing device to reassess and remap qubit assignments based on accumulated error data, improving execution quality while maintaining reasonable computational efficiency through scheduled rather than continuous transpilation operations.
Solution Approach 2:
The system implements feedback by using accumulated error information from circuit execution to inform subsequent transpilation decisions. The quantum processing device monitors qubit error rates and uses this feedback to adjust qubit mappings in later transpilation operations, creating a closed-loop system that continuously improves execution quality based on actual performance data.
2Device complexity
If automatic transpilation algorithm selects qubit mappings, then device complexity is reduced, but manufacturing precision deteriorates due to lack of fine-grained control
Solution Approach 1:
The system applies local quality by allowing different qubits to have different mapping strategies based on their individual error characteristics. Rather than applying a uniform automatic mapping algorithm to all qubits, the system identifies specific qubits with high error rates and applies targeted remapping strategies to those particular qubits, achieving fine-grained control where precision is most needed while keeping the overall process simple.
Solution Approach 2:
The system changes parameters by adjusting qubit mapping assignments based on measured error parameters. When qubit error rates exceed thresholds or change over time, the system modifies the mapping parameters to assign different virtual qubits to different physical qubits, thereby optimizing execution quality without requiring complete redesign of the transpilation process.
3Productivity
If fixed qubit mappings are used during multiple circuit shots, then transpilation computational overhead is reduced, but harmful factors increase due to accumulated circuit errors
Solution Approach 1:
The system implements periodic remapping during circuit execution by reassessing qubit mappings at predetermined intervals or after a certain number of circuit shots. This periodic intervention interrupts the accumulation of errors from fixed mappings while minimizing computational overhead by not performing continuous remapping, thus balancing execution speed with error mitigation.
Solution Approach 2:
The system introduces dynamics by making qubit mappings adaptable rather than static. The mapping configuration changes over time based on accumulated error data, allowing the system to transition from fixed mappings to dynamic remapping as execution progresses. This dynamic approach mitigates error accumulation while maintaining computational efficiency through conditional rather than constant remapping.
Data Source
AI summary
One example method includes obtaining error information for physical qubits that are candidates for mapping by respective virtual qubits of a quantum circuit, sampling the physical qubits based on their respective error information, mapping the virtual qubits to the physical qubits obtained by the sampling, and performing a shot of the quantum circuit on the sampled physical qubits. These operations may be performed ‘n’ times until an acceptable result is obtained for execution of the quantum circuit.