[Question] How to run approximate simulation in Python?

[rentristandelacruz]

How do I run approximate simulations (fidelity-driven or memory-driven) in Python?

I am mainly using DDSim's ReadTheDocs.io documentation for information.
(https://ddsim.readthedocs.io/en/latest/)

The CircuitSimulator page (https://ddsim.readthedocs.io/en/latest/simulators/CircuitSimulator.html) mentions that the Circuit Simulator is the “default” simulator in DDSIM with the following features: 1) exact simulation, 2) weak simulation, and 3) approximate simulation with different strategies. It has two sections "Usage in Python" and "Usage as Standalone Executable" but I can't find any instructions on how to do approximate simulations in Python.

For the standalone executable, I run "./ddsim_simple --help" and I can see that the following options can be set in order to run approximate simulations:
--step_fidelity
--steps arg
--approx_when
--approx_state

I can't seem to find any instructions on how to set this options when using Python.

[hillmich]

Thank you for bringing this issue to our attention. Indeed there was a problem with generating the docs for the bindings. You can have a look at #225.

For future reference, the signature for the __init__ method of the CircuitSimulator is the following:

__init__(self: mqt.ddsim.pyddsim.CircuitSimulator, circ: object, approximation_step_fidelity: float = 1.0, approximation_steps: int = 1, approximation_strategy: str = 'fidelity', seed: int = -1) → None

Does this help?

[rentristandelacruz]

I see. Thank you! I am able to run approximate simulations now in Python.

I'm not yet familiar with DDSim's code. I saw the approximation options (approximation_step_fidelity, approximation_steps, approximation_strategy) in mqt/ddsim/qasmsimulator.py.

I thought I can pass these options to the execute function like this:

backendSimulator = ddsim.DDSIMProvider().get_backend("qasm_simulator")
myJob = execute(quantumCircuit, backendSimulator, shots=10000, approximation_step_fidelity=0.8, approximation_steps=2)

I thought it was similar to passing the mode and nthreads options to the execute function when using the hybrid_qasm_simulator backend.

backendSimulator = ddsim.DDSIMProvider().get_backend("hybrid_qasm_simulator")
myJob = execute(quantumCircuit, backendSimulator, shots=10000, mode="dd", nthreads=2)

[hillmich]

The way you propose to use the approximation should work with the current version. The keyword paramaters are passed to the simulator.

The relevant snippet from qasmsimulator.py:

    def run_experiment(self, qobj_experiment: QasmQobjExperiment, **options) -> Dict:
        start_time = time.time()
        approximation_step_fidelity = options.get("approximation_step_fidelity", 1.0)
        approximation_steps = options.get("approximation_steps", 1)
        approximation_strategy = options.get("approximation_strategy", "fidelity")
        seed = options.get("seed", -1)

        sim = CircuitSimulator(
            qobj_experiment,
            approximation_step_fidelity=approximation_step_fidelity,
            approximation_steps=approximation_steps,
            approximation_strategy=approximation_strategy,
            seed=seed,
        )
        counts = sim.simulate(options.get("shots", 1024))
[...]

PR #225 adds some code to get rid of warnings and a simple test for the approximation. I'll add some documentation when I'll have time.

[rentristandelacruz]

I initially did not notice that passing those options (approximation_step_fidelity, approximation_steps, approximation_strategy) to execute function actually works because I was focusing on warnings like:

/home/user./local/lib/python3.10/site-packages/qiskit/execute_function.py:378: UserWarning: Option approximation_step_fidelity is not used by this backend
  job = backend.run(experiments, **run_kwargs)

After checking again, I finally noticed that changing the options (specifically I was changing the value of approximation_step_fidelity) actually changed the result (shots/counts) of the job/experiment even if the warnings are there.

Thanks!