v0.2.0 – Framework Improvements and Updated Examples

v0.2.0 - 2025-11-09

Added

• Complete NumPy-based optimizer suite under qmlhc.optim.numpy_optim/:
  adam, spsa, finite-diff, natural-grad, dual-ascent, mpc, kfac, and trust-kl.
• Unified optimizer factory and registry system create_optimizer_numpy() for flexible configuration.
• Added run_batch() method for consistent batched execution (B, D) across backends.
• Added a minimal optimizer usage example integrated with existing demo pipelines.
• Added backend examples and documentation for the capabilities() interface.

Changed

• Unified backend adapter layer for PennyLane, Qiskit, and Cpp backends with standardized conventions for
  capability reporting, batch execution, and noise handling.
• Added supports_noise keyword parameter to the PennyLane backend for explicit compatibility and control.
• Refined internal logic of pennylane_backend.qnode_shot for compatibility with the latest PennyLane versions (no API break).

Docs

• Added minimal optimizer documentation (table + quick example).
• Updated RST for qmlhc.backends and optimizer sections.
• Verified Sphinx build without critical warnings.

Testing

• Expanded unit and integration tests for optimizers, registry, and backend adapters.
• Achieved >99% overall test coverage with pytest --cov.

Removed

• Removed outdated figures and deprecated references (e.g., qmlhc.optim.api.rst).

Migration Notes

• No breaking changes introduced; existing APIs like run() remain functional.
• New parameters and features are optional and backward compatible.

---

[0.2.1] - (Planned)

Docs

• Extended optimizer documentation including parameter definitions, formulas, and detailed use cases.
• Added extended examples and Jupyter notebooks.

---