Skip to content

application_benchmarks.md

Latest commit

 

History

History
87 lines (68 loc) · 5.24 KB

application_benchmarks.md

File metadata and controls

87 lines (68 loc) · 5.24 KB

SPDX-License-Identifier: AGPL-3.0-or-later

Commercial license available

© Concepts 1996–2026 Miroslav Šotek. All rights reserved.

© Code 2020–2026 Miroslav Šotek. All rights reserved.

ORCID: 0009-0009-3560-0851

Contact: www.anulum.li | protoscience@anulum.li

scpn-quantum-control — Application benchmark plugins

Application Benchmark Plugins

Application plugins expose domain datasets through the same QPU data artifact contract used by the Kuramoto-XY pipeline. The built-in plugins cover EEG, tokamak MHD mode locking, IEEE power-grid synchronisation, and Friston-style predictive coding.

from scpn_quantum_control.applications import (
    compile_application_problem,
    run_application_benchmark_suite,
)

results = run_application_benchmark_suite()
problem = compile_application_problem("power_grid_ieee5")

Plugin Extras

The optional extras keep domain dependencies off the default install:

Extra Intended domain stack
app-eeg EEG/MEG file readers and MNE pipelines.
app-plasma HDF5/tabular tokamak or plasma diagnostics.
app-power-grid Power-system case readers and grid toolchains.
app-fep Structured predictive-coding workflow configuration.
app-benchmarks All four application stacks.

The packaged benchmark JSON files do not require those extras. The extras are for users who plug in external raw archives and want the same registry path to build a QPU-ready artifact.

Packaged Datasets

The in-repo artifacts live in data/public_application_benchmarks/. Each file validates as QPUDataArtifact, carries array hashes, and can be adapted to the public KuramotoProblem facade.

Dataset Plugin Pipeline path
eeg_alpha_plv_8ch eeg_alpha PLV matrix → QPU artifact → EEG topology benchmark → Kuramoto facade.
iter_mhd_8mode plasma_iter_mhd NTM/RWM mode graph → QPU artifact → mode-locking benchmark → Kuramoto facade.
ieee5bus_power_grid power_grid_ieee5 IEEE 5-bus constants → QPU artifact → grid synchronisation benchmark → Kuramoto facade.
friston_fep_6node friston_fep Precision graph + observations → QPU artifact → variational free energy + predictive-coding step.

Third-party plugins register factories under the scpn_quantum_control.application_plugins entry-point group. A broken plugin is logged and skipped so one domain adapter cannot block the rest of the benchmark suite.

Curated Researcher Workflows

The promoted researcher workflows are deliberately small and deterministic. They are meant to demonstrate the application boundary, provenance trail, and QPU-ready artefact format without presenting compact benchmark matrices as substitutes for raw domain archives.

Workflow Promoted artefacts Provenance boundary Deterministic regeneration
GraphML/CSV topology import External user-supplied graph or edge table converted to QPUDataArtifact. Bring-your-own topology path; the repository does not ship private third-party graph archives. Use the application-plugin registry and validate the converted artefact before adapting it to KuramotoProblem.
EEG alpha PLV data/public_application_benchmarks/eeg_alpha_plv_8ch.json; measured audit artefacts in data/knm_physical_validation/. Public-literature benchmark matrix for examples; raw EDF cohorts stay outside Git under .coordination/datasets/. scripts/build_real_eeg_plv_validation_dataset.py and scripts/compare_eeg_plv_cohorts.py regenerate the measured audit artefacts.
IEEE power grid data/public_application_benchmarks/ieee5bus_power_grid.json; data/knm_physical_validation/measured_couplings_power_grid_ieee5bus.json. Public IEEE 5-bus constants converted to swing-equation coupling; current measured-system comparison is a negative/control result. scripts/build_power_grid_measured_couplings.py regenerates the measured coupling and K_nm comparison artefacts.
Plasma/tokamak data/public_application_benchmarks/iter_mhd_8mode.json. Curated ITER-scale mode-locking topology from public MHD literature, not raw discharge traces. run_application_benchmark_suite() includes the packaged plasma benchmark without optional HDF5 dependencies.
Notebook and example workflows examples/02_kuramoto_xy_demo.py, examples/05_vqe_ansatz_comparison.py, examples/09_classical_vs_quantum_benchmark.py, examples/13_iter_disruption_demo.py, examples/18_end_to_end_pipeline.py, examples/19_sync_witness_operator.py, and examples/20_quantum_persistent_homology.py. Notebooks remain narrative wrappers; reusable logic stays in src/, scripts/, and versioned example files. Static example tests ensure promoted examples remain parseable, expose main(), and are listed in examples/README.md.

For a no-credential smoke path, run:

.venv-linux/bin/python - <<'PY'
from scpn_quantum_control.applications import run_application_benchmark_suite

results = run_application_benchmark_suite()
print(sorted(results))
PY

This command loads the packaged JSON artefacts only. It does not submit IBM jobs, download raw EEG data, or touch private datasets.