File size: 4,362 Bytes
67b79df b9d7591 67b79df | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 | # Dense-Evolution
**Dense statevector quantum simulator · JAX XLA · NISQ · VQE · QML**
[](https://github.com/tatopenn-cell/Dense-Evolution/actions/workflows/ci.yml)
[](https://pypi.org/project/dense-evolution/)
[](https://www.python.org/)
[](https://github.com/google/jax)
Dense-Evolution is a high-performance statevector simulator for deep NISQ circuits, VQE
pipelines, and QML workloads. It eliminates Kronecker-product overhead via stride-sliced
linear kernel fusion compiled through JAX XLA, keeping memory at the theoretical minimum
of `2ⁿ × 16 bytes`.
## What's in here
- **[`DenseSVSimulator`](api/simulator.md)** — the core dense statevector engine, JIT-fused
circuit execution, projective measurement.
- **[`MPSSimulator`](api/mps.md)** — matrix-product-state engine for low-entanglement
circuits at hundreds of qubits, with a `jax.lax.scan`-fused execution path.
- **[`Chunk`](api/chunk.md)** — anti-OOM statevector chunking for circuits too large for a
single dense allocation, including a real multi-device distributed dispatch path.
- **[`QASMParser`](api/parser.md)** — OpenQASM 2.0 / 3.0 parsing, including `for` loops,
range syntax, and QASM3 classical `bit` registers.
- **[`dense_evolution.mitigation`](api/mitigation.md)** — Zero-Noise Extrapolation, both the
classic scalar/vector form and a density-matrix extension (Smolin-Gambetta-Smith physical
projection, Uhlmann fidelity), every entry point with a `jax.jit`-compatible variant.
- **[`dense_evolution.healing`](api/healing.md)** — the predictive-healing primitives
(`calculate_delta_preemp`, `calculate_vettore_dinamico`, ...) ZNE's healing-adapted branch
is built on.
- **[`NoiseModel`](api/registry.md)** — Kraus-channel noise (depolarizing, bitflip,
phaseflip, amplitude damping, and a combined worst-case model), JAX- and NumPy-native.
- **Interop** with [Qiskit and PennyLane](api/interop.md), and [autodiff](api/autodiff.md)
through `circuit_to_energy_fn` for gradient-based VQE.
## Module dependencies
Real module graph, traced from the actual `import` statements in each file (not an
idealized layering) -- circuit sources feed three interchangeable execution engines, which
in turn feed the higher-level mitigation/autodiff layers.
```mermaid
flowchart LR
subgraph core["Core"]
parser["parser<br/>(QASMParser)"]
gates["gates"]
registry["registry<br/>(NoiseModel)"]
compiler["compiler<br/>(QuantumTranspiler)"]
end
subgraph engines["Engines"]
simulator["simulator<br/>(DenseSVSimulator)"]
mps["mps<br/>(MPSSimulator)"]
chunk["chunk<br/>(Chunk)"]
end
subgraph higher["Higher-level"]
interop["interop<br/>(Qiskit / PennyLane)"]
autodiff["autodiff<br/>(circuit_to_energy_fn)"]
healing["healing<br/>(predictive primitives)"]
mitigation["mitigation<br/>(ZNE)"]
end
gates --> registry
simulator --> registry
simulator --> gates
simulator --> compiler
mps --> compiler
mps --> gates
chunk --> simulator
chunk --> compiler
chunk --> gates
interop --> parser
interop --> simulator
autodiff --> parser
autodiff --> gates
autodiff --> compiler
autodiff --> registry
mitigation --> healing
```
## Where to start
New to the package? Start with **[Getting Started](getting-started.md)**.
Looking for a specific function or class? Jump straight to the **[API
Reference](api/index.md)**.
Want to see what changed recently? Check the **[Changelog](changelog.md)**.
## Honesty as a design principle
This library's docstrings and changelog document real, measured findings — including
negative results and rejected approaches — not just the features that worked. Where a
number is quoted (a speedup, a fidelity improvement, a coverage percentage), it was
measured directly and the measurement is described, not assumed. See the
[Changelog](changelog.md) for the full history.
|