---
title: Samplex inputs and outputs
description: Samplex inputs and outputs for the latest version of Samplomatic
source: https://eu-de.quantum.cloud.ibm.com/docs/en/addons/samplomatic/guides/samplex-io
---

# Samplex inputs and outputs

## Introduction

[`Samplex`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex "samplomatic.samplex.Samplex") is the core type of the samplomatic library. A samplex represents a parametric probability distribution over the parameters of some template circuit, with other array-valued fields to use when post-processing data collected from executing the bound template circuit. Its central interface is the [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample") method that draws from this distribution to produce a collection of arrays.

This guide describes how to use the [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample") interface, how to query and specify required inputs, and how to inspect expected outputs. The arrays supplied to and returned from [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample") are strongly typed. That is, for an instance of a samplex, their names, types, and shapes are fixed and queryable before sampling is done. The inputs and outputs present in a samplex depend on many factors. For example, whether measurement twirling is present and whether noise injection is required. For this reason, in general, two samplex instances are not expected to have the same inputs and outputs.

## Setup

First, construct a boxed-up circuit and build it into a samplex and template pair. These will be used in the examples that follow. Note that when manually boxing up the circuit, this guide uses the names `alpha`, `beta`, `ref1`, `ref2`, `mod_ref1`, `mod_ref2`, `mod_ref3`, and `conclude`. Additionally, the circuit has four `Parameter` objects. This guide shows how each of these play a role in the samplex inputs and outputs. See [the transpiler guide](/docs/addons/samplomatic/guides/transpiler) for information about boxing up circuits automatically.

```python
import matplotlib.pyplot as plt
import numpy as np
from qiskit.circuit import ClassicalRegister, Parameter, QuantumCircuit, QuantumRegister
from qiskit.quantum_info import Operator, Pauli, PauliLindbladMap

from samplomatic import ChangeBasis, InjectNoise, Twirl, build

# our circuit has two classical registers named alpha and beta
circuit = QuantumCircuit(
    QuantumRegister(4), alpha := ClassicalRegister(3, "alpha"), beta := ClassicalRegister(1, "beta")
)

# the first box is only twirled
with circuit.box([Twirl()]):
    for idx in range(4):
        circuit.rx(Parameter(f"a{idx}"), idx)
    circuit.cz(0, 1)
    circuit.cz(1, 2)

# the second box is twirled, and has noise injected
with circuit.box([Twirl(), InjectNoise(ref="ref1", modifier_ref="mod_ref1")]):
    circuit.h(1)
    circuit.cz(1, 2)

# the third box is twirled, and has different noise injected
with circuit.box([Twirl(), InjectNoise(ref="ref2", modifier_ref="mod_ref2")]):
    circuit.rx(0.1, range(4))
    circuit.cz(0, 1)
    circuit.cz(1, 2)

# the fourth box is the same as the second, but with a different modifer ref
with circuit.box([Twirl(), InjectNoise(ref="ref1", modifier_ref="mod_ref3")]):
    circuit.h(1)
    circuit.cz(1, 2)

circuit.barrier()

# the final two boxes twirl, and add a basis change in one case
with circuit.box([Twirl(), ChangeBasis(ref="conclude")]):
    circuit.measure(range(3), alpha)

with circuit.box([Twirl()]):
    circuit.measure([3], beta)

circuit.draw("mpl")
```

![../\_images/88f6576d7e2fd76bfb52d5133c6bc07e908a28941c6538f5251ac508cfb202f7.png](https://eu-de.quantum.cloud.ibm.com/docs/images/addons/samplomatic/88f6576d7e2fd76bfb52d5133c6bc07e908a28941c6538f5251ac508cfb202f7.avif)

Next, call [`build()`](/docs/api/samplomatic/auto/build#samplomatic.build "samplomatic.build") on the boxed circuit to construct a template and samplex pair. You will see how each of the boxes is turned into a barrier “sandwich” that includes `rz-sx-rz-sx-rz` fragments to implement dressing.

```python
template, samplex = build(circuit)

template.draw("mpl", fold=100)
```

![../\_images/ec8456d0d0a5928c5d056184cbd1842dcf7aefdc7cdba428f62092b27646d601.png](https://eu-de.quantum.cloud.ibm.com/docs/images/addons/samplomatic/ec8456d0d0a5928c5d056184cbd1842dcf7aefdc7cdba428f62092b27646d601.avif)

Plot the samplex DAG and hover over the nodes to see the following:

- All of the sampling nodes (stars) are responsible for generating randomizations for twirling, sampling from noise models, or injecting basis changes.
- All of the collection nodes (bow ties) are responsible for rendering output slices, which are measurement flips or parameter angles.
- All of the intermediate nodes (circles) are responsible for various kinds of data manipulation.

```python
samplex.draw()
```

## Query the required inputs and expected outputs

The easiest way to see the required inputs and expected outputs is to print the [`Samplex`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex "samplomatic.samplex.Samplex") object. Array items are formatted as `'{name}' <{type}[{shape}...]>`, and the required inputs precede the optional inputs.

```python
print(samplex)
```

```myst
Samplex(<58 nodes>)
  Inputs:
  - 'basis_changes.conclude' <uint8[3]>: Basis changing gates, in the symplectic ordering
      I=0, Z=1, X=2, and Y=3.
  - 'parameter_values' <float64[4]>: Input parameter values to use during sampling.
  - 'pauli_lindblad_maps.ref1' <PauliLindbladMap>: A PauliLindblad map acting on 2 qubits,
      with 'num_terms_ref1' terms.
  - 'pauli_lindblad_maps.ref2' <PauliLindbladMap>: A PauliLindblad map acting on 4 qubits,
      with 'num_terms_ref2' terms.

  - 'local_scales.mod_ref1' <float64['num_terms_ref1']>: (Optional) An array of factors by
      which to scale individual rates of a Pauli Lindblad map, where the order should match the
      corresponding list of terms.
  - 'local_scales.mod_ref2' <float64['num_terms_ref2']>: (Optional) An array of factors by
      which to scale individual rates of a Pauli Lindblad map, where the order should match the
      corresponding list of terms.
  - 'local_scales.mod_ref3' <float64['num_terms_ref1']>: (Optional) An array of factors by
      which to scale individual rates of a Pauli Lindblad map, where the order should match the
      corresponding list of terms.
  - 'noise_scales.mod_ref1' <float64[]>: (Optional) A scalar factor by which to scale rates
      of a Pauli Lindblad map.
  - 'noise_scales.mod_ref2' <float64[]>: (Optional) A scalar factor by which to scale rates
      of a Pauli Lindblad map.
  - 'noise_scales.mod_ref3' <float64[]>: (Optional) A scalar factor by which to scale rates
      of a Pauli Lindblad map.
  Outputs:
    * 'measurement_flips.alpha' <bool['num_randomizations', 1, 3]>: Bit-flip corrections
        for measurement twirling.
    * 'measurement_flips.beta' <bool['num_randomizations', 1, 1]>: Bit-flip corrections for
        measurement twirling.
    * 'parameter_values' <float32['num_randomizations', 48]>: Parameter values valid for an
        associated template circuit.
    * 'pauli_signs' <bool['num_randomizations', 3]>: Signs from sampled Pauli Lindblad
        maps, where boolean values represent the parity of the number of non-trivial factors in the
        sampled error that arise from negative rates. In other words, in order to implement basic
        PEC, the sign used to correct expectation values should be ``(-1)**bool_value``. The order
        matches the iteration order of boxes in the original circuit with noise injection
        annotations.
```

The [`inputs()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.inputs "samplomatic.samplex.Samplex.inputs") and [`outputs()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.outputs "samplomatic.samplex.Samplex.outputs") of the samplex, as well as more detailed information like their names, shapes, types, and descriptions, can also be queried programatically. The return type of both of these methods is a [`TensorInterface`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface "samplomatic.tensor_interface.TensorInterface"). For example, the following code lists all specifications of the input interface:

```python
samplex.inputs().specs
```

```myst
[TensorSpecification('basis_changes.conclude', (3,), dtype('uint8'), 'Basis changing gates, in the symplectic ordering I=0, Z=1, X=2, and Y=3.'),
 TensorSpecification('local_scales.mod_ref1', ('num_terms_ref1',), dtype('float64'), 'An array of factors by which to scale individual rates of a Pauli Lindblad map, where the order should match the corresponding list of terms.', optional=True),
 TensorSpecification('local_scales.mod_ref2', ('num_terms_ref2',), dtype('float64'), 'An array of factors by which to scale individual rates of a Pauli Lindblad map, where the order should match the corresponding list of terms.', optional=True),
 TensorSpecification('local_scales.mod_ref3', ('num_terms_ref1',), dtype('float64'), 'An array of factors by which to scale individual rates of a Pauli Lindblad map, where the order should match the corresponding list of terms.', optional=True),
 TensorSpecification('noise_scales.mod_ref1', (), dtype('float64'), 'A scalar factor by which to scale rates of a Pauli Lindblad map.', optional=True),
 TensorSpecification('noise_scales.mod_ref2', (), dtype('float64'), 'A scalar factor by which to scale rates of a Pauli Lindblad map.', optional=True),
 TensorSpecification('noise_scales.mod_ref3', (), dtype('float64'), 'A scalar factor by which to scale rates of a Pauli Lindblad map.', optional=True),
 TensorSpecification('parameter_values', (4,), dtype('float64'), 'Input parameter values to use during sampling.'),
 PauliLindbladMapSpecification('pauli_lindblad_maps.ref1', num_qubits=2, num_terms=num_terms_ref1),
 PauliLindbladMapSpecification('pauli_lindblad_maps.ref2', num_qubits=4, num_terms=num_terms_ref2)]
```

The next example prints some information about those output specifiers whose names contain the string `"flips"`.

```python
for spec in samplex.outputs().get_specs("flips"):
    print(spec.name, spec.shape, spec.dtype)
```

```myst
measurement_flips.alpha ('num_randomizations', 1, 3) bool
measurement_flips.beta ('num_randomizations', 1, 1) bool
```

## Bind input values and sample

To [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample"), the samplex needs values for at least those [`inputs()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.inputs "samplomatic.samplex.Samplex.inputs") that are marked as required. These values are first bound to an input interface so that it can perform all necessary type checking (or type coercion in some cases) and shape analysis. If anything is wrong with the types or shapes, or if a required value is missing, a verbose error is raised.

### Bind input

In the following example, the [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample") method is passed the required input values. It is then used to draw three randomizations from the samplex. The required bindings from the dressings are the Pauli Lindblad maps for noise injection and the basis change array. Additionally, values for the four parameters in the original boxed circuit are required, which in this case is set to `np.linspace(0, 1, 4)`. The samplex composes these parameters into the dressings, so they effectively become modifications to the samplex’s output parameter values.

```python
inputs = samplex.inputs().bind(
    # bind() concatenates nested dicts with a '.' so that we can
    # set 'pauli_lindblad_maps.noise0/1' as follows. note that the
    # names 'ref1' and 'ref2' derive from the names in the InjectNoise
    # annotation.
    pauli_lindblad_maps={
        "ref1": PauliLindbladMap.identity(2),
        "ref2": PauliLindbladMap.identity(4),
    },
    # likewise, we can set all basis changes as follows, where
    # the name 'conclude' comes from our basis change annotation
    basis_changes={"conclude": [0, 1, 2]},
    # we must provide values for all parameters in the original circuit
    parameter_values=np.linspace(0, 1, 4),
)

outputs = samplex.sample(inputs, num_randomizations=3)
print(outputs)
```

```myst
SamplexOutput(<
  - Dimension constraints: num_randomizations=3

  * 'measurement_flips.alpha' <bool['num_randomizations', 1, 3]>: Bit-flip corrections for
      measurement twirling.
  * 'measurement_flips.beta' <bool['num_randomizations', 1, 1]>: Bit-flip corrections for
      measurement twirling.
  * 'parameter_values' <float32['num_randomizations', 48]>: Parameter values valid for an
      associated template circuit.
  * 'pauli_signs' <bool['num_randomizations', 3]>: Signs from sampled Pauli Lindblad maps,
      where boolean values represent the parity of the number of non-trivial factors in the sampled
      error that arise from negative rates. In other words, in order to implement basic PEC, the
      sign used to correct expectation values should be ``(-1)**bool_value``. The order matches the
      iteration order of boxes in the original circuit with noise injection annotations.
>)
```

Inputs can be bound to the interface through multiple calls to [`bind()`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface.bind "samplomatic.tensor_interface.TensorInterface.bind") or by direct item assignment. Despite the nested dictionary format that [`bind()`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface.bind "samplomatic.tensor_interface.TensorInterface.bind") allows for convenience, it is a flat mapping object, and supports all of the usual mapping syntax. Once all required inputs have been found, [`fully_bound`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface.fully_bound "samplomatic.tensor_interface.TensorInterface.fully_bound") becomes true.

```python
inputs = (
    samplex.inputs()
    .bind(
        pauli_lindblad_maps={
            "ref1": PauliLindbladMap.identity(2),
            "ref2": PauliLindbladMap.identity(4),
        }
    )
    .bind(parameter_values=np.linspace(0, 1, 4))
)

# once the final requirement is bound, fully_bound becomes True
print("Fully bound?", inputs.fully_bound)
inputs["basis_changes.conclude"] = [0, 2, 3]
print("Fully bound?", inputs.fully_bound)

# setting optional values does not affect whether the interface is fully bound
inputs["noise_scales.mod_ref1"] = 3
print("Fully bound?", inputs.fully_bound)
```

```myst
Fully bound? False
Fully bound? True
Fully bound? True
```

All [`TensorInterface`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface "samplomatic.tensor_interface.TensorInterface") objects, including the [`SamplexOutput`](/docs/api/samplomatic/auto/samplex-samplex-output#samplomatic.samplex.SamplexOutput "samplomatic.samplex.SamplexOutput") subclass, map objects against data that has been bound to them. This means that you can get, for example, the template parameter values as follows. In this case, it has shape `(3, 48)` since three randomizations were requested and the template circuit has 48 parameters.

```python
print("(num_randomizations, num_template_params) =", outputs["parameter_values"].shape)
print(outputs["parameter_values"][0, :])
```

```myst
(num_randomizations, num_template_params) = (3, 48)
[ 0.0000000e+00  0.0000000e+00  0.0000000e+00  1.5707964e+00
  3.3333334e-01  1.5707964e+00 -1.5707964e+00  2.4749260e+00
 -1.5707964e+00  1.5707964e+00  1.0000000e+00  1.5707964e+00
  0.0000000e+00  1.5707964e+00  0.0000000e+00  3.1415927e+00
  0.0000000e+00  0.0000000e+00  1.5707964e+00  3.0415926e+00
 -1.5707964e+00 -1.5707964e+00  3.0415926e+00 -1.5707964e+00
 -1.5707964e+00  1.0000000e-01  1.5707964e+00  1.5707964e+00
  1.0000000e-01  1.5707964e+00  3.1415927e+00  1.5707964e+00
  0.0000000e+00  0.0000000e+00  0.0000000e+00  0.0000000e+00
  0.0000000e+00  3.1415927e+00  0.0000000e+00  3.1415927e+00
  0.0000000e+00  0.0000000e+00  3.1415927e+00  1.5707964e+00
  0.0000000e+00  4.4408921e-16  3.1415927e+00  0.0000000e+00]
```

### Specify input as a dictionary

Instead of providing a fully-bound [`TensorInterface`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface "samplomatic.tensor_interface.TensorInterface"), you can specify the inputs as a standard dictionary, as is shown in the cell below. In this case, the [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample") method internally binds the dictionary items to the samplex’s input interface to validate that all required inputs are present and compatible.

```python
inputs = {
    "pauli_lindblad_maps": {
        "ref1": PauliLindbladMap.identity(2),
        "ref2": PauliLindbladMap.identity(4),
    },
    "basis_changes": {"conclude": [0, 1, 2]},
    "parameter_values": np.linspace(0, 1, 4),
}
outputs = samplex.sample(inputs, num_randomizations=3)
```

### Input types

Currently, there are two interface value types:

- Array-valued: These have a shape and a data type. Bound values are coerced into the data type if possible. Any dimension can be a free dimension.
- Pauli Lindblad maps: These must be of type [`qiskit.quantum_info.PauliLindbladMap`](/docs/api/qiskit/qiskit.quantum_info.PauliLindbladMap) and must act on the prescribed number of qubits. The number of terms in the map is a free dimension.

```python
# up to precision loss, these are equivalent via automatic type coersion of array inputs
samplex.inputs()["parameter_values"] = [1, 2, 8, 9]
samplex.inputs()["parameter_values"] = np.array([1, 2, 8, 9], dtype=np.float16)
```

### Free dimensions

The [`TensorInterface`](/docs/api/samplomatic/auto/tensor-interface-tensor-interface#samplomatic.tensor_interface.TensorInterface "samplomatic.tensor_interface.TensorInterface") class has a notion of *free dimensions*, which are named integer sizes whose values are undetermined until some data has been bound to the interface that constrains them. All free dimensions of the same name must resolve to a consistent size when binding data, or an error will be raised. For example, in the [`outputs()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.outputs "samplomatic.samplex.Samplex.outputs") of any samplex, the array axis corresponding to randomizations is a free parameter that has the name `"num_randomizations"` by convention. When `sample()` is called, it binds values to the outputs

```python
print("All free dimensions:", outputs.free_dimensions)
print("Current constraints:", outputs.bound_dimensions)
```

```myst
All free dimensions: {'num_randomizations'}
Current constraints: {'num_randomizations': 3}
```

Another common free-dimension scenario is sharing term constraints between Pauli lindblad map specifications and their modifiers. A [`qiskit.quantum_info.PauliLindbladMap`](/docs/api/qiskit/qiskit.quantum_info.PauliLindbladMap) is an ordered sequence of terms, where each term contains a floating point rate and a sparse Pauli representation. The [`build()`](/docs/api/samplomatic/auto/build#samplomatic.build "samplomatic.build") function adds a specifier `"pauli_lindblad_maps.<ref>"` for this type for each [`InjectNoise`](/docs/api/samplomatic/auto/inject-noise#samplomatic.InjectNoise "samplomatic.InjectNoise") annotation, and if it contains a `modifier_ref`, a `"local_scales.<modifier_ref>"` specifier is also added. Both of these share the free dimension `"num_terms_<ref>"` that specifies how many Pauli Lindblad terms are present in the noise model because the latter needs to zip against the rates of the former to scale them.

```python
# both the PauliLindbladMap and the local scales imply 2 terms, so that the free dimension
# 'num_terms_noise2' is satisfied
inputs = samplex.inputs().bind(
    pauli_lindblad_maps={"ref2": PauliLindbladMap.from_list([("XXYZ", 0.1), ("IIXX", 0.2)])},
    local_scales={"mod_ref2": [1, 2]},
)
print("All free dimensions:", inputs.free_dimensions)
print("Current constraints:", inputs.bound_dimensions)
```

```myst
All free dimensions: {'num_terms_ref1', 'num_terms_ref2'}
Current constraints: {'num_terms_ref2': 2}
```

## Qubit ordering convention

Samplexes constructed from boxed circuits by the [`build()`](/docs/api/samplomatic/auto/build#samplomatic.build "samplomatic.build") function can require array-valued inputs to [`sample()`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex.sample "samplomatic.samplex.Samplex.sample"), where one index corresponds to qubits in a circuit’s box. This section explains the ordering convention that should be used for such axes. In summary: use the qubit order of the boxed circuit, restricted to the box’s qubits.

For example, suppose you want to figure out how to specify a Pauli Lindblad map for `"noise1"`. Then the order of qubits in the noise map is with respect to these qubits:

```python
box_of_interest = circuit[1]
qubit_ordering_convention = [qubit for qubit in circuit.qubits if qubit in box_of_interest.qubits]
qubit_ordering_convention
```

```myst
[<Qubit register=(4, "q0"), index=1>, <Qubit register=(4, "q0"), index=2>]
```

That is, if you want `<Qubit register=(4, "q0"), index=2>` to be noisy with `X` noise, then you should define a Pauli Lindblad map like this one because it is at index `1` of `qubit_ordering_convention`.

```python
samplex.inputs().bind(
    pauli_lindblad_maps={
        "ref1": (noise1 := PauliLindbladMap.from_sparse_list([("X", [1], 0.12)], num_qubits=2))
    }
)
noise1
```

```myst
<PauliLindbladMap with 1 term on 2 qubits: (0.12)L(X_1)>
```

The same holds true for basis changes. If you want to rotate into the +1 eigenstates of the operators X, Y, and Z for qubits 0, 1, and 2 of `circuit`, respectively, then you should use the following array, recalling that the convention $I=0$, $Z=1$, $X=2$, and $Y=3$ is used in this library (and Qiskit).

```python
samplex.inputs().bind(basis_changes={"conclude": (basis_change := [2, 3, 1])})
basis_change
```

```myst
[2, 3, 1]
```

Note that the qubit ordering convention is *not* the order defined by either of the following:

- `box_instruction.qubits`, where `box_instruction` is a [`qiskit.circuit.CircuitInstruction`](/docs/api/qiskit/qiskit.circuit.CircuitInstruction) whose operation is a [`qiskit.circuit.BoxOp`](/docs/api/qiskit/qiskit.circuit.BoxOp), because the order of this list of qubits might depend on the order in which instructions were added to the box context, which would be a confusing convention. For example, `with box(): circuit.x([0, 1])` and `with box(): circuit.x([1, 0])` would result in different orderings.
- `box_instruction.operation.body.qubits` because these qubits might not even appear in the outer circuit. Qubits inside the actual body of `BoxOp` instructions (or any other control flow operation like `IfElseOp`) are only promised to have a consisent meaning within that scope. This would be an ill-defined convention, and hard to use in general.

## Local testing with the template circuit

If a [`Samplex`](/docs/api/samplomatic/auto/samplex-samplex#samplomatic.samplex.Samplex "samplomatic.samplex.Samplex") is constructed by using the [`build()`](/docs/api/samplomatic/auto/build#samplomatic.build "samplomatic.build") function, the output parameter values should be directly compatible with the parameters of the associated template circuit. For example, you can bind values for one of the randomizations to the template for local inspection.

```python
inputs = samplex.inputs().bind(
    pauli_lindblad_maps={
        "ref1": PauliLindbladMap.identity(2),
        "ref2": PauliLindbladMap.identity(4),
    },
    basis_changes={"conclude": [0, 0, 0]},
    parameter_values=np.linspace(0, 1, 4),
)

outputs = samplex.sample(inputs, num_randomizations=3)

bound_template = template.assign_parameters(outputs["parameter_values"][2])
bound_template.draw("mpl", fold=1000)
```

![../\_images/1d82c46209f8648a2df572cfef8ca256f47952566d34386761ab942aa6e50afd.png](https://eu-de.quantum.cloud.ibm.com/docs/images/addons/samplomatic/1d82c46209f8648a2df572cfef8ca256f47952566d34386761ab942aa6e50afd.avif)

This lets you verify that the samplex is outputing samples that are expected. For example, the following code casts the bound circuit to a [`qiskit.quantum_info.Operator`](/docs/api/qiskit/qiskit.quantum_info.Operator) object and composes with the bitflip Paulis to visually inspect that all three randomizations are logically equivalent to the base circuit.

```python
from samplomatic.utils import unbox

# Operator requires that we unpack all of the boxes first
unboxed = unbox(circuit).assign_parameters(inputs["parameter_values"])
# We want to plot unitary matrix amplitudes, so we need to remove the non-unitary measurements
unboxed.remove_final_measurements()
unboxed_unitary = Operator(unboxed)

# Plot magnitudes of the unitary matrix represenation of the circuit
plt.subplot(1, 4, 1)
plt.imshow(np.abs(unboxed_unitary))
plt.title("Base Circuit")

# For each randomization we sampled, plot the unitary matrix representation
for idx in range(3):
    bound_template = template.assign_parameters(outputs["parameter_values"][idx])
    bound_template.remove_final_measurements()

    # The example does measurement twirling by compiling random bitflip gates before the
    # measurements, so we need to undo these at the operator level for each randomization.
    # We do this by casting them to Paulis and composing with the bound circuit's unitary
    alpha_flips = Pauli(([0, 0, 0], outputs["measurement_flips.alpha"][idx, 0]))
    beta_flips = Pauli(([0], outputs["measurement_flips.beta"][idx, 0]))
    flips = beta_flips ^ alpha_flips
    bound_unitary = Operator(bound_template) & flips

    # Plot magnitudes of the unitary matrix represenation of the circuit
    plt.subplot(1, 4, idx + 2)
    plt.imshow(np.abs(bound_unitary))
    plt.title(f"Randomization {idx}")

plt.tight_layout()
```

![../\_images/af1aa74f8a0e8abed673d4cd43ff8b810c44297d2f3a42fcfb49ade811e51c58.png](https://eu-de.quantum.cloud.ibm.com/docs/images/addons/samplomatic/af1aa74f8a0e8abed673d4cd43ff8b810c44297d2f3a42fcfb49ade811e51c58.avif)

Similarly, the following code passes output parameter values and the template to a sampler to simulate 10,000 shots of all randomizations, and compares the resulting expectation values (only of the alpha register) to the expectation values of a base circuit simulation.

```python
from qiskit.primitives import StatevectorEstimator as Estimator
from qiskit.primitives import StatevectorSampler as Sampler
from qiskit.primitives.containers import BitArray

# simulate some expectation values direction from the base circuit
estimator_job = Estimator().run([(unboxed, ["IZII", "IIZI", "IIIZ"])])
evs = estimator_job.result()[0].data["evs"]

# get the sampler data from each randomization, and do bitflip correction
sampler_job = Sampler().run([(template, outputs["parameter_values"])], shots=10_000)
alpha_data = sampler_job.result()[0].data["alpha"]
alpha_data ^= BitArray.from_bool_array(outputs["measurement_flips.alpha"], "little")

# compare the results
print("  EVs from base circuit:", evs)
print("EVs from randomizations:", alpha_data.expectation_values(["ZII", "IZI", "IIZ"]))
```

```myst
  EVs from base circuit: [0.7819611  0.93553883 0.99500417]
EVs from randomizations: [0.789  0.9366 0.995 ]
```