qml.measurements¶
Warning
Unless you are a PennyLane or plugin developer, you likely do not need to use these classes directly.
See the main measurements page for details on available measurements.
This module contains all the measurements supported by PennyLane.
Description¶
Measurements¶
The MeasurementProcess
class serves as a base class for measurements, and is inherited
from by the SampleMeasurement
, StateMeasurement
and MeasurementTransform
classes. These classes are subclassed to implement measurements in PennyLane.
Each
SampleMeasurement
subclass represents a samplebased measurement, which contains aSampleMeasurement.process_samples()
method and aSampleMeasurement.process_counts()
method that process the sequence of samples generated by the device.process_samples
method should always have the same arguments:samples (Sequence[complex]): computational basis samples generated for all wires
wire_order (Wires): wires determining the subspace that
samples
acts onshot_range (tuple[int]): 2tuple of integers specifying the range of samples to use. If not specified, all samples are used.
bin_size (int): Divides the shot range into bins of size
bin_size
, and returns the measurement statistic separately over each bin. If not provided, the entire shot range is treated as a single bin.
SampleMeasurement.process_counts()
is currently optional. It accepts a dictionary mapping a string representation of a basis state to an integer and a wire order.See
CountsMP
for an example.Each
StateMeasurement
subclass represents a statebased measurement, which contains aStateMeasurement.process_state()
method that processes the quantum state generated by the device. This method should always have the same arguments:state (Sequence[complex]): quantum state
wire_order (Wires): wires determining the subspace that
state
acts on; a matrix of dimension \(2^n\) acts on a subspace of \(n\) wires
See
StateMP
for an example.Each
MeasurementTransform
subclass represents a measurement process that requires the application of a batch transform, which contains aMeasurementTransform.process()
method that converts the given quantum tape into a batch of quantum tapes and executes them using the device. This method should always have the same arguments:tape (QuantumTape): quantum tape to transform
device (Device): device used to transform the quantum tape
The main difference between a
MeasurementTransform
and abatch_transform()
is that a batch transform is tracked by the gradient transform, while aMeasurementTransform
process isn’t.See
ClassicalShadowMP
for an example.
Note
A measurement process can inherit from both SampleMeasurement
and
StateMeasurement
classes, defining the needed logic to process either samples or the
quantum state. See VarianceMP
for an example.
Differentiation¶
In general, a MeasurementProcess
is differentiable with respect to a parameter if the domain of
that parameter is continuous. When using the analytic method of differentiation the output of the
measurement process must be a real scalar value for it to be differentiable.
Working with midcircuit measurements¶
Midcircuit measurements can be made using qml.measure()
. The measurement value is returned by qml.measure
and can be used as a condition for classical control. Moreover, multiple measurement values can be combined
using arithmetic operators for more complex conditioning:
import pennylane as qml
dev = qml.device("default.qubit", wires=3)
@qml.qnode(dev)
def circ(x, y):
qml.RX(x, wires=0)
qml.RY(y, wires=1)
m0 = qml.measure(0)
m1 = qml.measure(1)
qml.cond(~m0 & m1 == 0, qml.X)(wires=2)
return qml.expval(qml.Z(2))
Wires can be reused as normal after making midcircuit measurements. Moreover, a measured wire can also be
reset to the \(0 \rangle\) state by setting the reset
keyword argument of qml.measure
to True
.
Users can also collect statistics on midcircuit measurements along with other terminal measurements. Currently,
qml.expval
, qml.probs
, qml.sample
, qml.counts
, and qml.var
are supported. Users have the
ability to collect statistics on single measurement values.
import pennylane as qml
dev = qml.device("default.qubit", wires=3)
@qml.qnode(dev)
def circ(x, y):
qml.RX(x, wires=0)
qml.RY(y, wires=1)
m0 = qml.measure(1)
return qml.expval(qml.Z(0)), qml.sample(m0)
QNodes can be executed as usual when collecting midcircuit measurement statistics:
>>> circ(1.0, 2.0, shots=5)
(0.6, array([1, 1, 1, 0, 1]))
PennyLane also supports postselecting on midcircuit measurement outcomes. To learn more, refer to the documentation
of measure()
.
Creating custom measurements¶
A custom measurement process can be created by inheriting from any of the classes mentioned above.
The following is an example for a samplebased measurement that computes the number of samples obtained of a given state:
import pennylane as qml
from pennylane.measurements import SampleMeasurement
class CountState(SampleMeasurement):
def __init__(self, state: str):
self.state = state # string identifying the state e.g. "0101"
wires = list(range(len(state)))
super().__init__(wires=wires)
def process_samples(self, samples, wire_order, shot_range=None, bin_size=None):
counts_mp = qml.counts(wires=self._wires)
counts = counts_mp.process_samples(samples, wire_order, shot_range, bin_size)
return float(counts.get(self.state, 0))
def process_counts(self, counts, wire_order):
return float(counts.get(self.state, 0))
def __copy__(self):
return CountState(state=self.state)
Note
The __copy__
method needs to be overriden when new arguments are added into the __init__
method.
The measurement process in this example uses the counts()
function, which is a
measurement process which returns a dictionary containing the number of times each quantum
state has been sampled.
We can now execute the new measurement in a QNode
. Let’s use a simple circuit
so that we can verify our results mathematically.
dev = qml.device("default.qubit", wires=1, shots=10000)
@qml.qnode(dev)
def circuit(x):
qml.RX(x, wires=0)
return CountState(state="1")
The quantum state before the measurement will be:
When \(\theta = 1.23\), the probability of obtaining the state \(\begin{bmatrix} 1 \\ 0 \end{bmatrix}\) is \(\sin^2(\theta/2) = 0.333\). Using 10000 shots we should obtain the excited state 3333 times approximately.
>>> circuit(1.23)
array(3303.)
Given that the measurement process returns a real scalar value, we can differentiate it using the analytic method.
We know from the previous analysis that the analytic result of the measurement process is \(r(\theta) = \text{nshots} \cdot \sin^2(\theta/2)\).
The gradient of the measurement process is \(\frac{\partial r}{\partial \theta} = \text{nshots} \sin(\theta/2) \cos(\theta/2)\).
When \(\theta = 1.23\), \(\frac{\partial r}{\partial \theta} = 4712.444\)
>>> x = qml.numpy.array(1.23, requires_grad=True)
>>> qml.grad(circuit)(x)
4715.000000000001
Note
In PennyLane we use functions to define measurements (e.g. counts()
). These
functions will return an instance of the corresponding measurement process
(e.g. CountsMP
). This decision is just for design purposes.
Serialization and Pytree format
PennyLane measurements are automatically registered as Pytrees .
MeasurementProcess._flatten
and MeasurementProcess._unflatten
need to be overwritten if the measurement has additional
metadata, such as seed
or all_outcomes
.
>>> H = 2.0 * qml.X(0)
>>> mp = qml.expval(H)
>>> mp._flatten()
((<Hamiltonian: terms=1, wires=[0]>, None), ())
>>> type(mp)._unflatten(*mp._flatten())
expval( (2) [X0])
>>> jax.tree_util.tree_leaves(mp)
[2]
Adding your new measurement to PennyLane¶
If you want to add this new measurement to PennyLane such that other users can benefit from it, you have to make a Pull Request
creating a file with the name of your measurement (e.g. state.py
) and add it in
pennylane/measurements/
. This file should contain:
The measurement class with the appropriate
process
method defined.A function with the same name as the created file that will be used to instantiate the measurement class.
The tests are added to a file of a similar name and location in tests/measurements/
.
The class and the function need to be imported in pennylane/measurements/__init__.py
. The
function needs to be imported in pennylane/__init__.py
.
Here are a few more tips for adding measurements:
Choose the name carefully. Good names tell the user what the measurement is used for, or what architecture it implements. Ask yourself if a measurement of a similar name could be added soon in a different context.
Write good docstrings. Explain what your measurement does in a clear docstring with ample examples.
You can find more about Pennylane standards in the guidelines on Documentation.
Overview¶
Toplevel measurement functions¶
The measurements
functions for creating standard PennyLane measurement
processes are toplevel imports:

The classical shadow measurement protocol. 

Sample from the supplied observable, with the number of shots determined from the 

Quantum density matrix in the computational basis. 

Expectation value of the supplied observable. 

Perform a midcircuit measurement in the computational basis on the supplied qubit. 

Mutual information between the subsystems prior to measurement: 

Probability of each computational basis state. 

The purity of the system prior to measurement. 

Sample from the supplied observable, with the number of shots determined from the 

Compute expectation values using classical shadows in a differentiable manner. 

Quantum state in the computational basis. 

Variance of the supplied observable. 

Von Neumann entropy of the system prior to measurement. 
Functions¶
Classes¶

Represents a classical shadow measurement process occurring at the end of a quantum variational circuit. 

Measurement process that samples from the supplied observable and returns the number of counts for each sample. 

Measurement process that returns the quantum state in the computational basis. 

Measurement process that computes the expectation value of the supplied observable. 

Represents a measurement process occurring at the end of a quantum variational circuit. 

Measurement process that applies a transform into the given quantum tape. 

A class representing unknown measurement outcomes in the qubit model. 

Midcircuit measurement. 

Measurement process that computes the mutual information between the provided wires. 

Measurement process that computes the probability of each computational basis state. 

Measurement process that computes the purity of the system prior to measurement. 

Measurement process that returns the samples of a given observable. 

Samplebased measurement process. 

Measures the expectation value of an operator using the classical shadow measurement process. 

A namedtuple that represents a shot quantity being repeated some number of times. 

A data class that stores shot information. 

Measurement process that returns the quantum state in the computational basis. 

Statebased measurement process. 

Measurement process that computes the variance of the supplied observable. 

Measurement process that computes the Von Neumann entropy of the system prior to measurement. 
Variables¶
Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 

Enumeration class to represent the return types of an observable. 