tn-noise-learning

Learn the noise channel of a quantum circuit layer from measurement shots, with tensor networks.

A quantum channel on n qubits is stored as a locally purified density operator (LPDO): one Kraus tensor V_j per qubit with a Kraus index of size K, joined to its neighbours by virtual bonds of size χ along a chosen Topology. The channel Λ(ρ) = Σ V ρ V† is completely positive by construction. Its parameters are fitted by maximum likelihood to shots of product-state preparations and product measurements, with a soft penalty that keeps the channel trace preserving.

Topology  ─►  LPDO  ─►  NoiseDataset (zarr)  ─►  Trainer  ─►  History.plot()

• Installation: how to install the package.
• Quickstart: the example script, step by step.
• Concepts: the LPDO, the probability network, the loss, how shots are simulated.
• Developer guide: layout, conventions, how to extend.
• Contributing: development environment, everyday commands, pull requests.
• References: the papers this package builds on.
• Code reference: generated from the docstrings.