# Techniques: Approximation Methods for Simulation and Equivalence Checking of Noisy Quantum Circuits

## What the paper contributes (one paragraph)
The paper gives a classical approximation framework for simulating and checking the equivalence of noisy quantum circuits, aimed at scaling these tasks to NISQ circuits with hundreds of qubits under low noise. A noisy circuit is encoded as a double-size tensor network using the matrix representation of each noise channel, and singular value decomposition is applied to each noise matrix to obtain a low-rank decomposition whose dominant term approximates the channel. Substituting noises with their dominant term, then progressively reintroducing residual terms, yields an l-level approximation with provable error bounds and controllable cost. The same machinery is reused for approximate equivalence checking via the Jamiolkowski fidelity between an ideal unitary and a noisy super-operator. Implemented on Google's TensorNetwork package, the method simulates and equivalence-checks QAOA circuits with around 200 qubits and 20 noise operators, outperforming the quantum trajectories method and MPO-based simulation in speed and scalability.

## Techniques offered
- **Tensor-network classical simulation of noisy circuits with SVD low-rank noise approximation** - represents a noisy circuit as a double-size (2n-qubit) tensor network and approximates each noise channel by the dominant singular term of its matrix representation, supplying a deterministic estimate of a measurement probability with a tunable accuracy/cost trade-off.
  - guarantee: statistical error bound on the estimated probability, |G - G'| < sum_{u=l+1}^{N} C(N,u) (16p)^u (1+16p)^{N-u}, with contraction count O(15^l N^l) (Theorem 1)
  - quote: "To handle the difficulty of simulating a circuit with a large number of noises, we introduce an approximation noisy circuit simulation method based on the matrix representation and SVD, which can balance the accuracy and efficiency of the simulation."
- **Approximate equivalence checking of noisy circuits via Jamiolkowski fidelity** - reuses the SVD tensor-network approximation to estimate the Jamiolkowski fidelity between an ideal unitary channel and a noisy super-operator, supplying a quantitative degree-of-equivalence check between an ideal circuit and its noisy implementation.
  - guarantee: error bound on the fidelity estimate, |F_J(U, E_N) - F(l)| <= sum_{u=l+1}^{N} C(N,u) (16p)^u (1+16p)^{N-u}, with contraction count O(15^l N^l) (Theorem 2)
  - quote: "Based on this representation, our approximation technique in Section~\ref{sec:approximation_algorithm} yields a practical equivalence-checking algorithm; see Algorithm~\ref{Algorithm:ApproximateEC}."

## Where it could apply
- Target primitive(s): QAOA, VQE, QFT, Trotterised and general gate-based circuits, sampling / random (supremacy) circuits; any noisy circuit whose noise is given as a super-operator
- Target application group(s): platforms_tooling, quantum_ml, error_correction (circuit verification and noise-impact assessment), chemistry_materials (VQE)
- Code family / hardware assumptions: hardware-agnostic; near-identity (low-rate) single-qubit and two-qubit noise channels, including superconducting decoherence and crosstalk models (e.g. ibm_brisbane topology); no code family assumed

## Caveats
Accuracy holds only in the low-noise regime: the error bound and the advantage over the quantum trajectories method degrade as the noise rate p grows, and the method (like its baselines) fails on circuits with very many noise operators that inject substantial entanglement. The guarantees are numerical error and resource bounds on a classical estimator, not interactive verification or a certificate executed on quantum hardware. Equivalence checking is restricted to the approximate (fidelity) setting with one channel taken to be an ideal unitary, not exact equivalence.