Noise Modeling
A noise model describes how errors affect the results from executing a quantum circuit. Noise models support several stages of a typical workflow including: simulating a circuit classically to estimate expected performance before submitting a job or to inform an error-mitigation technique such as probabilistic error cancellation. For general strategies to reduce the impact of noise during execution, see Overview of noise management techniques.
This page collects the packages and capabilities that support each part of that workflow, grouped into learning noise from a device, building a noise model for classical simulation, and refining a noise model for a specific error-mitigation technique.
Building and running noise learning algorithms
Use noise learning packages to characterize a device's noise directly from experimental data. These packages run randomized circuits to estimate error rates and build a noise model that reflects actual hardware behavior.
Build randomization-based quantum noise characterization experiments and model-fitting analysis pipelines
Browse documentation
Construct noise models for PEC and PEA workloads using a target device with the IBM Quantum Compute service.
Browse documentation
Building noise models for noisy circuit simulation
Use this package to construct a noise model for simulating circuits classically. With a noise model, you can estimate how a circuit behaves on noisy hardware before you run it on a QPU.
Refining noise models
Use these packages to adjust a noise model for a specific error-mitigation technique. Each package targets a different tradeoff between sampling cost, bias, and the reliability of postselected samples.
Reduce probabilistic error cancellation (PEC) sampling overhead by removing noise model terms that have low impact on the observable estimation.
Browse documentation
Reduce probabilistic error cancellation (PEC) sampling overhead by removing noise model terms that can be detected by error checks in the circuit.
Browse documentation