Skip to content

v0.2.5

Choose a tag to compare

View all tags
@FedorShind FedorShind released this 24 Mar 18:56
· 76 commits to main since this release
v0.2.5
b27af9e

What's new

Layerwise Richardson ZNE

Circuits with heterogeneous layer structure -- where multi-qubit gate density varies across layers -- now get a tailored ZNE configuration. When layer heterogeneity exceeds 2.0 and depth is in the 15-50 range, EMRG recommends RichardsonFactory with fold_gates_at_random instead of fold_global. This targets noise amplification at the most error-prone gates rather than scaling uniformly.

  • New layer_heterogeneity field in circuit analysis output
  • Automatic layerwise detection -- no new flags needed
  • Visible in emrg analyze (table and JSON) and in generated code comments

Tutorial notebooks

Two complete, runnable Jupyter notebooks in docs/tutorials/:

  • VQE H2 mitigation -- 4-qubit hardware-efficient ansatz, walks through analysis, ZNE recipe generation, and PEC comparison
  • QAOA MaxCut mitigation -- 5-node graph with 2-layer QAOA, shows how deeper circuits get different EMRG recommendations

Bug fixes

  • recommend() now raises ValueError for invalid technique strings instead of silently falling through to ZNE
  • Fixed OverflowError in PEC overhead estimation on circuits with 200+ multi-qubit gates

Benchmarks

Updated benchmark suite with four parts: tool performance, ZNE fidelity, PEC vs ZNE head-to-head comparison, and layerwise vs global folding on high-heterogeneity circuits. PEC benchmarks now use meaningful test circuits and measure both single-qubit and multi-qubit observables.

Testing

  • 284 tests, 99% coverage
  • New permanent stress test suite (tests/stress_test.py) covering edge cases, boundary conditions, and performance regression

Full changelog

  • feat(analyzer): add layer heterogeneity metric
  • feat(heuristics): add layerwise Richardson ZNE variant
  • feat(codegen): support layerwise ZNE code generation
  • feat(cli): show layer heterogeneity in analyze output
  • docs: add VQE H2 and QAOA MaxCut tutorial notebooks
  • fix: validate technique parameter in recommend()
  • fix: prevent OverflowError in PEC overhead estimation
  • bench: update benchmarks for v0.2.5 with PEC and layerwise results
  • bench: improve PEC and layerwise benchmark circuits
Assets 2
Loading