Skip to main contentIBM Quantum Documentation

Qiskit 0.23 release notes

0.23.6

Terra 0.16.4

No change

Aer 0.7.5

Prelude

This release is a bugfix release that fixes compatibility in the precompiled binary wheel packages with numpy versions < 1.20.0. The previous release 0.7.4 was building the binaries in a way that would require numpy 1.20.0 which has been resolved now, so the precompiled binary wheel packages will work with any numpy compatible version.

Ignis 0.5.2

No change

Aqua 0.8.2

No change

IBM Q Provider 0.11.1

No change

0.23.5

Terra 0.16.4

Prelude

This release is a bugfix release that primarily fixes compatibility with numpy 1.20.0. This numpy release deprecated their local aliases for Python’s numeric types (np.int -> int, np.float -> float, etc.) and the usage of these aliases in Qiskit resulted in a large number of deprecation warnings being emitted. This release fixes this so you can run Qiskit with numpy 1.20.0 without those deprecation warnings.

Aer 0.7.4

Bug Fixes

Fixes compatibility with numpy 1.20.0. This numpy release deprecated their local aliases for Python’s numeric types (np.int -> int, np.float -> float, etc.) and the usage of these aliases in Qiskit Aer resulted in a large number of deprecation warnings being emitted. This release fixes this so you can run Qiskit Aer with numpy 1.20.0 without those deprecation warnings.

Ignis 0.5.2

Prelude

This release is a bugfix release that primarily fixes compatibility with numpy 1.20.0. It is also the first release to include support for Python 3.9. Earlier releases (including 0.5.0 and 0.5.1) worked with Python 3.9 but did not indicate this in the package metadata, and there was no upstream testing for those releases. This release fixes that and was tested on Python 3.9 (in addition to 3.6, 3.7, and 3.8).

Bug Fixes

  • networkx(opens in a new tab) is explicitly listed as a dependency now. It previously was an implicit dependency as it was required for the qiskit.ignis.verification.topological_codes module but was not correctly listed as a depdendency as qiskit-terra also requires networkx and is also a depdency of ignis so it would always be installed in practice. However, it is necessary to list it as a requirement for future releases of qiskit-terra that will not require networkx. It’s also important to correctly list the dependencies of ignis in case there were a future incompatibility between version requirements.

Aqua 0.8.2

IBM Q Provider 0.11.1

No change

0.23.4

Terra 0.16.3

Bug Fixes

  • Fixed an issue introduced in 0.16.2 that would cause errors when running transpile() on a circuit with a series of 1 qubit gates and a non-gate instruction that only operates on a qubit (e.g. Reset). Fixes #5736(opens in a new tab)

Aer 0.7.3

No change

Ignis 0.5.1

No change

Aqua 0.8.1

No change

IBM Q Provider 0.11.1

No change

0.23.3

Terra 0.16.2

New Features

  • Python 3.9 support has been added in this release. You can now run Qiskit Terra using Python 3.9.

Upgrade Notes

  • The class MCXGrayCode will now create a C3XGate if num_ctrl_qubits is 3 and a C4XGate if num_ctrl_qubits is 4. This is in addition to the previous functionality where for any of the modes of the :class:’qiskit.library.standard_gates.x.MCXGate`, if num_ctrl_bits is 1, a CXGate is created, and if 2, a CCXGate is created.

Bug Fixes

  • Pulse Delay instructions are now explicitly assembled as PulseQobjInstruction objects included in the PulseQobj output from assemble().

    Previously, we could ignore Delay instructions in a Schedule as part of assemble() as the time was explicit in the PulseQobj objects. But, now with pulse gates, there are situations where we can schedule ONLY a delay, and not including the delay itself would remove the delay.

  • Circuits with custom gate calibrations can now be scheduled with the transpiler without explicitly providing the durations of each circuit calibration.

  • The BasisTranslator and Unroller passes, in some cases, had not been preserving the global phase of the circuit under transpilation. This has been fixed.

  • A bug in qiskit.pulse.builder.frequency_offset() where when compensate_phase was set a factor of 2π2\pi was missing from the appended phase.

  • Fix the global phase of the output of the QuantumCircuit method repeat(). If a circuit with global phase is appended to another circuit, the global phase is currently not propagated. Simulators rely on this, since the phase otherwise gets applied multiple times. This sets the global phase of repeat() to 0 before appending the repeated circuit instead of multiplying the existing phase times the number of repetitions.

  • Fixes bug in SparsePauliOp where multiplying by a certain non Python builtin Numpy scalar types returned incorrect values. Fixes #5408(opens in a new tab)

  • The definition of the Hellinger fidelity from has been corrected from the previous defition of 1−H(P,Q)1-H(P,Q) to [1−H(P,Q)2]2[1-H(P,Q)^2]^2 so that it is equal to the quantum state fidelity of P, Q as diagonal density matrices.

  • Reduce the number of CX gates in the decomposition of the 3-controlled X gate, C3XGate. Compiled and optimized in the U CX basis, now only 14 CX and 16 U gates are used instead of 20 and 22, respectively.

  • Fixes the issue wherein using Jupyter backend widget or qiskit.tools.backend_monitor() would fail if the backend’s basis gates do not include the traditional u1, u2, and u3.

  • When running qiskit.compiler.transpile() on a list of circuits with a single element, the function used to return a circuit instead of a list. Now, when qiskit.compiler.transpile() is called with a list, it will return a list even if that list has a single element. See #5260(opens in a new tab).

    from qiskit import *
 
qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
qc.measure_all()
 
transpiled = transpile([qc])
print(type(transpiled), len(transpiled))
    <class 'list'> 1

Aer 0.7.3

New Features

  • Python 3.9 support has been added in this release. You can now run Qiskit Aer using Python 3.9 without building from source.

Bug Fixes

  • Fixes issue with setting QasmSimulator basis gates when using "method" and "noise_model" options together, and when using them with a simulator constructed using from_backend(). Now the listed basis gates will be the intersection of gates supported by the backend configuration, simulation method, and noise model basis gates. If the intersection of the noise model basis gates and simulator basis gates is empty a warning will be logged.
  • Fixes a bug that resulted in c_if not working when the width of the conditional register was greater than 64. See #1077(opens in a new tab).
  • Fixes bug in from_backend() and from_backend() where basis_gates was set incorrectly for IBMQ devices with basis gate set ['id', 'rz', 'sx', 'x', 'cx']. Now the noise model will always have the same basis gates as the backend basis gates regardless of whether those instructions have errors in the noise model or not.
  • Fixes a bug when applying truncation in the matrix product state method of the QasmSimulator.

Ignis 0.5.1

No change

Aqua 0.8.1

No change

IBM Q Provider 0.11.1

No change

0.23.2

Terra 0.16.1

No change

Aer 0.7.2

New Features

  • Add the CMake flag DISABLE_CONAN (default=``OFF``)s. When installing from source, setting this to ON allows bypassing the Conan package manager to find libraries that are already installed on your system. This is also available as an environment variable DISABLE_CONAN, which takes precedence over the CMake flag. This is not the official procedure to build AER. Thus, the user is responsible of providing all needed libraries and corresponding files to make them findable to CMake.

Bug Fixes

  • Fixes a bug with nested OpenMP flag was being set to true when it shouldn’t be.

Ignis 0.5.1

No change

Aqua 0.8.1

No change

IBM Q Provider 0.11.1

No change

0.23.1

Terra 0.16.1

Bug Fixes

  • Fixed an issue where an error was thrown in execute for valid circuits built with delays.
  • The QASM definition of ‘c4x’ in qelib1.inc has been corrected to match the standard library definition for C4XGate.
  • Fixes a bug in subtraction for quantum channels A−BA - B where BB was an Operator object. Negation was being applied to the matrix in the Operator representation which is not equivalent to negation in the quantum channel representation.
  • Changes the way _evolve_instruction() access qubits to handle the case of an instruction with multiple registers.

Aer 0.7.1

Upgrade Notes

  • The minimum cmake version to build qiskit-aer has increased from 3.6 to 3.8. This change was necessary to enable fixing GPU version builds that support running on x86_64 CPUs lacking AVX2 instructions.

Bug Fixes

  • qiskit-aer with GPU support will now work on systems with x86_64 CPUs lacking AVX2 instructions. Previously, the GPU package would only run if the AVX2 instructions were available. Fixes #1023(opens in a new tab)
  • Fixes bug with AerProvider where options set on the returned backends using set_options() were stored in the provider and would persist for subsequent calls to get_backend() for the same named backend. Now every call to and backends() returns a new instance of the simulator backend that can be configured.
  • Fixes bug in the error message returned when a circuit contains unsupported simulator instructions. Previously some supported instructions were also being listed in the error message along with the unsupported instructions.
  • Fix bug where the “sx”` gate SXGate was not listed as a supported gate in the C++ code, in StateOpSet of matrix_product_state.hp.
  • Fix bug where "csx", "cu2", "cu3" were incorrectly listed as supported basis gates for the "density_matrix" method of the QasmSimulator.
  • In MPS, apply_kraus was operating directly on the input bits in the parameter qubits, instead of on the internal qubits. In the MPS algorithm, the qubits are constantly moving around so all operations should be applied to the internal qubits.
  • When invoking MPS::sample_measure, we need to first sort the qubits to the default ordering because this is the assumption in qasm_controller.This is done by invoking the method move_all_qubits_to_sorted_ordering. It was correct in sample_measure_using_apply_measure, but missing in sample_measure_using_probabilities.

Ignis 0.5.1

Bug Fixes

  • Fix the "auto" method of the TomographyFitter, StateTomographyFitter, and ProcessTomographyFitter to only use "cvx" if CVXPY is installed and a third-party SDP solver other than SCS is available. This is because the SCS solver has lower accuracy than other solver methods and often returns a density matrix or Choi-matrix that is not completely-positive and fails validation when used with the qiskit.quantum_info.state_fidelity() or qiskit.quantum_info.process_fidelity() functions.

Aqua 0.8.1

0.8.1

New Features

  • A new algorithm has been added: the Born Openheimer Potential Energy surface for the calculation of potential energy surface along different degrees of freedom of the molecule. The algorithm is called BOPESSampler. It further provides functionalities of fitting the potential energy surface to an analytic function of predefined potentials.some details.

Critical Issues

  • Be aware that initial_state parameter in QAOA has now different implementation as a result of a bug fix. The previous implementation wrongly mixed the user provided initial_state with Hadamard gates. The issue is fixed now. No attention needed if your code does not make use of the user provided initial_state parameter.

Bug Fixes

  • optimize_svm method of qp_solver would sometimes fail resulting in an error like this ValueError: cannot reshape array of size 1 into shape (200,1) This addresses the issue by adding an L2 norm parameter, lambda2, which defaults to 0.001 but can be changed via the QSVM algorithm, as needed, to facilitate convergence.
  • A method one_letter_symbol has been removed from the VarType in the latest build of DOCplex making Aqua incompatible with this version. So instead of using this method an explicit type check of variable types has been introduced in the Aqua optimization module.
  • :meth`~qiskit.aqua.operators.state_fns.DictStateFn.sample()` could only handle real amplitudes, but it is fixed to handle complex amplitudes. #1311 <https://github.com/Qiskit/qiskit-aqua/issues/1311(opens in a new tab)> for more details.
  • Trotter class did not use the reps argument in constructor. #1317 <https://github.com/Qiskit/qiskit-aqua/issues/1317(opens in a new tab)> for more details.
  • Raise an AquaError if :class`qiskit.aqua.operators.converters.CircuitSampler` samples an empty operator. #1321 <https://github.com/Qiskit/qiskit-aqua/issues/1321(opens in a new tab)> for more details.
  • to_opflow() returns a correct operator when coefficients are complex numbers. #1381 <https://github.com/Qiskit/qiskit-aqua/issues/1381(opens in a new tab)> for more details.
  • Let backend simulators validate NoiseModel support instead of restricting to Aer only in QuantumInstance.
  • Correctly handle PassManager on QuantumInstance transpile method by calling its run method if it exists.
  • A bug that mixes custom initial_state in QAOA with Hadamard gates has been fixed. This doesn’t change functionality of QAOA if no initial_state is provided by the user. Attention should be taken if your implementation uses QAOA with cusom initial_state parameter as the optimization results might differ.
  • Previously, setting seed_simulator=0 in the QuantumInstance did not set any seed. This was only affecting the value 0. This has been fixed.

IBM Q Provider 0.11.1

New Features

  • qiskit.providers.ibmq.experiment.Experiment now has three additional attributes, hub, group, and project, that identify the provider used to create the experiment.
  • Methods qiskit.providers.ibmq.experiment.ExperimentService.experiments() and qiskit.providers.ibmq.experiment.ExperimentService.analysis_results() now support a limit parameter that allows you to limit the number of experiments and analysis results returned.

Upgrade Notes

  • A new parameter, limit is now the first parameter for both qiskit.providers.ibmq.experiment.ExperimentService.experiments() and qiskit.providers.ibmq.experiment.ExperimentService.analysis_results() methods. This limit has a default value of 10, meaning by deafult only 10 experiments and analysis results will be returned.

Bug Fixes

  • Fixes the issue wherein a job could be left in the CREATING state if job submit fails half-way through.
  • Fixes the infinite loop raised when passing an IBMQRandomService instance to a child process.

0.23.0

Terra 0.16.0

Prelude

The 0.16.0 release includes several new features and bug fixes. The major features in this release are the following:

  • Introduction of scheduled circuits, where delays can be used to control the timing and alignment of operations in the circuit.
  • Compilation of quantum circuits from classical functions, such as oracles.
  • Ability to compile and optimize single qubit rotations over different Euler basis as well as the phase + square-root(X) basis (i.e. ['p', 'sx']), which will replace the older IBM Quantum basis of ['u1', 'u2', 'u3'].
  • Tracking of global_phase() on the QuantumCircuit class has been extended through the transpiler, quantum_info, and assembler modules, as well as the BasicAer and Aer simulators. Unitary and state vector simulations will now return global phase-correct unitary matrices and state vectors.

Also of particular importance for this release is that Python 3.5 is no longer supported. If you are using Qiskit Terra with Python 3.5, the 0.15.2 release is that last version which will work.

New Features

  • Global R gates have been added to qiskit.circuit.library. This includes the global R gate (GR), global Rx (GRX) and global Ry (GRY) gates which are derived from the GR gate, and global Rz ( GRZ) that is defined in a similar way to the GR gates. The global R gates are defined on a number of qubits simultaneously, and act as a direct sum of R gates on each qubit.

    For example:

    from qiskit import QuantumCircuit, QuantumRegister
import numpy as np
 
num_qubits = 3
qr = QuantumRegister(num_qubits)
qc = QuantumCircuit(qr)
 
qc.compose(GR(num_qubits, theta=np.pi/3, phi=2*np.pi/3), inplace=True)

    will create a QuantumCircuit on a QuantumRegister of 3 qubits and perform a RGate of an angle θ=π3\theta = \frac{\pi}{3} about an axis in the xy-plane of the Bloch spheres that makes an angle of ϕ=2π3\phi = \frac{2\pi}{3} with the x-axis on each qubit.

  • A new color scheme, iqx, has been added to the mpl backend for the circuit drawer qiskit.visualization.circuit_drawer() and qiskit.circuit.QuantumCircuit.draw(). This uses the same color scheme as the Circuit Composer on the IBM Quantum Experience website. There are now 3 available color schemes - default, iqx, and bw.

    There are two ways to select a color scheme. The first is to use a user config file, by default in the ~/.qiskit directory, in the file settings.conf under the [Default] heading, a user can enter circuit_mpl_style = iqx to select the iqx color scheme.

    The second way is to add {'name': 'iqx'} to the style kwarg to the QuantumCircuit.draw method or to the circuit_drawer function. The second way will override the setting in the settings.conf file. For example:

    from qiskit.circuit import QuantumCircuit
 
circuit = QuantumCircuit(2)
circuit.h(0)
circuit.cx(0, 1)
circuit.measure_all()
circuit.draw('mpl', style={'name': 'iqx'})

  • In the style kwarg for the the circuit drawer qiskit.visualization.circuit_drawer() and qiskit.circuit.QuantumCircuit.draw() the displaycolor field with the mpl backend now allows for entering both the gate color and the text color for each gate type in the form (gate_color, text_color). This allows the use of light and dark gate colors with contrasting text colors. Users can still set only the gate color, in which case the gatetextcolor field will be used. Gate colors can be set in the style dict for any number of gate types, from one to the entire displaycolor dict. For example:

    from qiskit.circuit import QuantumCircuit
 
circuit = QuantumCircuit(1)
circuit.h(0)
 
style_dict = {'displaycolor': {'h': ('#FA74A6', '#000000')}}
circuit.draw('mpl', style=style_dict)

    or

    style_dict = {'displaycolor': {'h': '#FA74A6'}}
circuit.draw('mpl', style=style_dict)

  • Two alignment contexts are added to the pulse builder (qiskit.pulse.builder) to facilitate writing a repeated pulse sequence with delays.

    • qiskit.pulse.builder.align_equispaced() inserts delays with equivalent length in between pulse schedules within the context.
    • qiskit.pulse.builder.align_func() offers more advanced control of pulse position. This context takes a callable that calculates a fractional coordinate of i-th pulse and aligns pulses within the context. This makes coding of dynamical decoupling easy.

  • A rep_delay parameter has been added to the QasmQobj class under the run configuration, QasmQobjConfig. This parameter is used to denote the time between program executions. It must be chosen from the backend range given by the BackendConfiguration method rep_delay_range(). If a value is not provided a backend default, qiskit.providers.models.BackendConfiguration.default_rep_delay, will be used. rep_delay will only work on backends which allow for dynamic repetition time. This is can be checked with the BackendConfiguration property dynamic_reprate_enabled.

  • The qobj_schema.json JSON Schema file in qiskit.schemas has been updated to include the rep_delay as an optional configuration property for QASM Qobjs.

  • The backend_configuration_schema.json JSON Schema file in qiskit.schemas has been updated to include dynamic_reprate_enabled, rep_delay_range and default_rep_delay as optional properties for a QASM backend configuration payload.

  • A new optimization pass, qiskit.transpiler.passes.TemplateOptimization has been added to the transpiler. This pass applies a template matching algorithm described in arXiv:1909.05270(opens in a new tab) that replaces all compatible maximal matches in the circuit.

    To implement this new transpiler pass a new module, template_circuits, was added to the circuit library (qiskit.circuit.library). This new module contains all the Toffoli circuit templates used in the TemplateOptimization.

    This new pass is not currently included in the preset pass managers (qiskit.transpiler.preset_passmanagers), to use it you will need to create a custom PassManager.

  • A new version of the providers interface has been added. This new interface, which can be found in qiskit.providers, provides a new versioning mechanism that will enable changes to the interface to happen in a compatible manner over time. The new interface should be simple to migrate existing providers, as it is mostly identical except for the explicit versioning.

    Besides having explicitly versioned abstract classes the key changes for the new interface are that the BackendV1 method run() can now take a QuantumCircuit or Schedule object as inputs instead of Qobj objects. To go along with that options are now part of a backend class so that users can configure run time options when running with a circuit. The final change is that qiskit.providers.JobV1 can now be synchronous or asynchronous, the exact configuration and method for configuring this is up to the provider, but there are interface hook points to make it explicit which execution model a job is running under in the JobV1 abstract class.

  • A new kwarg, inplace, has been added to the function qiskit.result.marginal_counts(). This kwarg is used to control whether the contents are marginalized in place or a new copy is returned, for Result object input. This parameter does not have any effect for an input dict or Counts object.

  • An initial version of a classical function compiler, qiskit.circuit.classicalfunction, has been added. This enables compiling typed python functions (operating only on bits of type Int1 at the moment) into QuantumCircuit objects. For example:

    from qiskit.circuit import classical_function, Int1
 
@classical_function
def grover_oracle(a: Int1, b: Int1, c: Int1, d: Int1) -> Int1:
     x = not a and b
     y = d and not c
     z = not x or y
     return z
 
quantum_circuit = grover_oracle.synth()
quantum_circuit.draw()

    The parameter registerless=False in the qiskit.circuit.classicalfunction.ClassicalFunction method synth() creates a circuit with registers refering to the parameter names. For example:

    quantum_circuit = grover_oracle.synth(registerless=False)
quantum_circuit.draw()

    A decorated classical function can be used the same way as any other quantum gate when appending it to a circuit.

    circuit = QuantumCircuit(5)
circuit.append(grover_oracle, range(5))
circuit.draw()

    The GROVER_ORACLE gate is synthesized when its decomposition is required.

    circuit.decompose().draw()

    The feature requires tweedledum, a library for synthesizing quantum circuits, that can be installed via pip with pip install tweedledum.

  • A new class qiskit.circuit.Delay for representing a delay instruction in a circuit has been added. A new method delay() is now available for easily appending delays to circuits. This makes it possible to describe timing-sensitive experiments (e.g. T1/T2 experiment) in the circuit level.

    from qiskit import QuantumCircuit
 
qc = QuantumCircuit(1, 1)
qc.delay(500, 0, unit='ns')
qc.measure(0, 0)
 
qc.draw()

  • A new argument scheduling_method for qiskit.compiler.transpile() has been added. It is required when transpiling circuits with delays. If scheduling_method is specified, the transpiler returns a scheduled circuit such that all idle times in it are padded with delays (i.e. start time of each instruction is uniquely determined). This makes it possible to see how scheduled instructions (gates) look in the circuit level.

    from qiskit import QuantumCircuit, transpile
from qiskit.test.mock.backends import FakeAthens
 
qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
 
scheduled_circuit = transpile(qc, backend=FakeAthens(), scheduling_method="alap")
print("Duration in dt:", scheduled_circuit.duration)
scheduled_circuit.draw(idle_wires=False)

    See also timeline_drawer() for the best visualization of scheduled circuits.

  • A new fuction qiskit.compiler.sequence() has been also added so that we can convert a scheduled circuit into a Schedule to make it executable on a pulse-enabled backend.

    from qiskit.compiler import sequence
 
sched = sequence(scheduled_circuit, pulse_enabled_backend)

  • The schedule() has been updated so that it can schedule circuits with delays. Now there are two paths to schedule a circuit with delay:

    qc = QuantumCircuit(1, 1)
qc.h(0)
qc.delay(500, 0, unit='ns')
qc.h(0)
qc.measure(0, 0)
 
sched_path1 = schedule(qc.decompose(), backend)
sched_path2 = sequence(transpile(qc, backend, scheduling_method='alap'), backend)
assert pad(sched_path1) == sched_path2

    Refer to the release notes and documentation for transpile() and sequence() for the details on the other path.

  • Added the GroverOperator to the circuit library (qiskit.circuit.library) to construct the Grover operator used in Grover’s search algorithm and Quantum Amplitude Amplification/Estimation. Provided with an oracle in form of a circuit, GroverOperator creates the textbook Grover operator. To generalize this for amplitude amplification and use a generic operator instead of Hadamard gates as state preparation, the state_in argument can be used.

  • The InstructionScheduleMap methods get() and pop() methods now take ParameterExpression instances in addition to numerical values for schedule generator parameters. If the generator is a function, expressions may be bound before or within the function call. If the generator is a ParametrizedSchedule, expressions must be bound before the schedule itself is bound/called.

  • A new class LinearAmplitudeFunction was added to the circuit library (qiskit.circuit.library) for mapping (piecewise) linear functions on qubit amplitudes,

    F∣x⟩∣0⟩=1−f(x)∣x⟩∣0⟩+f(x)∣x⟩∣1⟩F|x\rangle |0\rangle = \sqrt{1 - f(x)}|x\rangle |0\rangle + \sqrt{f(x)}|x\rangle |1\rangle

    The mapping is based on a controlled Pauli Y-rotations and a Taylor approximation, as described in https://arxiv.org/abs/1806.06893(opens in a new tab). This circuit can be used to compute expectation values of linear functions using the quantum amplitude estimation algorithm.

  • The new jupyter magic monospaced_output has been added to the qiskit.tools.jupyter module. This magic sets the Jupyter notebook output font to “Courier New”, when possible. When used this fonts returns text circuit drawings that are better aligned.

    import qiskit.tools.jupyter
%monospaced_output

  • A new transpiler pass, Optimize1qGatesDecomposition, has been added. This transpiler pass is an alternative to the existing Optimize1qGates that uses the OneQubitEulerDecomposer class to decompose and simplify a chain of single qubit gates. This method is compatible with any basis set, while Optimize1qGates only works for u1, u2, and u3. The default pass managers for optimization_level 1, 2, and 3 have been updated to use this new pass if the basis set doesn’t include u1, u2, or u3.

  • The OneQubitEulerDecomposer now supports two new basis, 'PSX' and 'U'. These can be specified with the basis kwarg on the constructor. This will decompose the matrix into a circuit using PGate and SXGate for 'PSX', and UGate for 'U'.

  • A new method remove() has been added to the qiskit.transpiler.PassManager class. This method enables removing a pass from a PassManager instance. It works on indexes, similar to replace(). For example, to remove the RemoveResetInZeroState pass from the pass manager used at optimization level 1:

    from qiskit.transpiler.preset_passmanagers import level_1_pass_manager
from qiskit.transpiler.passmanager_config import PassManagerConfig
 
pm = level_1_pass_manager(PassManagerConfig())
pm.draw()
    [0] FlowLinear: UnrollCustomDefinitions, BasisTranslator
[1] FlowLinear: RemoveResetInZeroState
[2] DoWhile: Depth, FixedPoint, Optimize1qGates, CXCancellation

    The stage [1] with RemoveResetInZeroState can be removed like this:

    pass_manager.remove(1)
pass_manager.draw()
    [0] FlowLinear: UnrollCustomDefinitions, BasisTranslator
[1] DoWhile: Depth, FixedPoint, Optimize1qGates, CXCancellation

  • Several classes to load probability distributions into qubit amplitudes; UniformDistribution, NormalDistribution, and LogNormalDistribution were added to the circuit library (qiskit.circuit.library). The normal and log-normal distribution support both univariate and multivariate distributions. These circuits are central to applications in finance where quantum amplitude estimation is used.

  • Support for pulse gates has been added to the QuantumCircuit class. This enables a QuantumCircuit to override (for basis gates) or specify (for standard and custom gates) a definition of a Gate operation in terms of time-ordered signals across hardware channels. In other words, it enables the option to provide pulse-level custom gate calibrations.

    The circuits are built exactly as before. For example:

    from qiskit import pulse
from qiskit.circuit import QuantumCircuit, Gate
 
class RxGate(Gate):
    def __init__(self, theta):
        super().__init__('rxtheta', 1, [theta])
 
circ = QuantumCircuit(1)
circ.h(0)
circ.append(RxGate(3.14), [0])

    Then, the calibration for the gate can be registered using the QuantumCircuit method add_calibration() which takes a Schedule definition as well as the qubits and parameters that it is defined for:

    # Define the gate implementation as a schedule
with pulse.build() as custom_h_schedule:
    pulse.play(pulse.library.Drag(...), pulse.DriveChannel(0))
 
with pulse.build() as q1_x180:
    pulse.play(pulse.library.Gaussian(...), pulse.DriveChannel(1))
 
# Register the schedule to the gate
circ.add_calibration('h', [0], custom_h_schedule)  # or gate.name string to register
circ.add_calibration(RxGate(3.14), [0], q1_x180)   # Can accept gate

    Previously, this functionality could only be used through complete Pulse Schedules. Additionally, circuits can now be submitted to backends with your custom definitions (dependent on backend support).

    Circuits with pulse gates can still be lowered to a Schedule by using the schedule() function.

    The calibrated gate can also be transpiled using the regular transpilation process:

    transpiled_circuit = transpile(circ, backend)

    The transpiled circuit will leave the calibrated gates on the same qubit as the original circuit and will not unroll them to the basis gates.

  • Support for disassembly of PulseQobj objects has been added to the qiskit.assembler.disassemble() function. For example:

    from qiskit import pulse
from qiskit.assembler.disassemble import disassemble
from qiskit.compiler.assemble import assemble
from qiskit.test.mock import FakeOpenPulse2Q
 
backend = FakeOpenPulse2Q()
 
d0 = pulse.DriveChannel(0)
d1 = pulse.DriveChannel(1)
with pulse.build(backend) as sched:
    with pulse.align_right():
        pulse.play(pulse.library.Constant(10, 1.0), d0)
        pulse.shift_phase(3.11, d0)
        pulse.measure_all()
 
qobj = assemble(sched, backend=backend, shots=512)
scheds, run_config, header = disassemble(qobj)

  • A new kwarg, coord_type has been added to qiskit.visualization.plot_bloch_vector(). This kwarg enables changing the coordinate system used for the input parameter that describes the positioning of the vector on the Bloch sphere in the generated visualization. There are 2 supported values for this new kwarg, 'cartesian' (the default value) and 'spherical'. If the coord_type kwarg is set to 'spherical' the list of parameters taken in are of the form [r, theta,  phi] where r is the radius, theta is the inclination from +z direction, and phi is the azimuth from +x direction. For example:

    from numpy import pi
 
from qiskit.visualization import plot_bloch_vector
 
x = 0
y = 0
z = 1
r = 1
theta = pi
phi = 0
 
 
# Cartesian coordinates, where (x,y,z) are cartesian coordinates
# for bloch vector
plot_bloch_vector([x,y,z])
    plot_bloch_vector([x,y,z], coord_type="cartesian")  # Same as line above
    # Spherical coordinates, where (r,theta,phi) are spherical coordinates
# for bloch vector
plot_bloch_vector([r, theta, phi], coord_type="spherical")

  • Pulse Schedule objects now support using ParameterExpression objects for parameters.

    For example:

    from qiskit.circuit import Parameter
from qiskit import pulse
 
alpha = Parameter('⍺')
phi = Parameter('ϕ')
qubit = Parameter('q')
amp = Parameter('amp')
 
schedule = pulse.Schedule()
schedule += SetFrequency(alpha, DriveChannel(qubit))
schedule += ShiftPhase(phi, DriveChannel(qubit))
schedule += Play(Gaussian(duration=128, sigma=4, amp=amp),
                 DriveChannel(qubit))
schedule += ShiftPhase(-phi, DriveChannel(qubit))

    Parameter assignment is done via the assign_parameters() method:

    schedule.assign_parameters({alpha: 4.5e9, phi: 1.57,
                            qubit: 0, amp: 0.2})

    Expressions and partial assignment also work, such as:

    beta = Parameter('b')
schedule += SetFrequency(alpha + beta, DriveChannel(0))
schedule.assign_parameters({alpha: 4.5e9})
schedule.assign_parameters({beta: phi / 6.28})

  • A new visualization function timeline_drawer() was added to the qiskit.visualization module.

    For example:

    from qiskit.visualization import timeline_drawer
from qiskit import QuantumCircuit, transpile
from qiskit.test.mock import FakeAthens
 
qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0,1)
timeline_drawer(transpile(qc, FakeAthens(), scheduling_method='alap'))

Upgrade Notes

  • Type checking for the params kwarg of the constructor for the Gate class and its subclasses has been changed. Previously all Gate parameters had to be in a set of allowed types defined in the Instruction class. Now a new method, validate_parameter() is used to determine if a parameter type is valid or not. The definition of this method in a subclass will take priority over its parent. For example, UnitaryGate accepts a parameter of the type numpy.ndarray and defines a custom validate_parameter() method that returns the parameter if it’s an numpy.ndarray. This takes priority over the function defined in its parent class Gate. If UnitaryGate were to be used as parent for a new class, this validate_parameter method would be used unless the new child class defines its own method.

  • The previously deprecated methods, arguments, and properties named n_qubits and numberofqubits have been removed. These were deprecated in the 0.13.0 release. The full set of changes are:

    ClassOldNew
    QuantumCircuitn_qubitsnum_qubits
    Paulinumberofqubitsnum_qubits
    FunctionOld ArgumentNew Argument
    qiskit.circuit.random.random_circuit()n_qubitsnum_qubits
    qiskit.circuit.library.MSGaten_qubitsnum_qubits

  • Inserting a parameterized Gate instance into a QuantumCircuit now creates a copy of that gate which is used in the circuit. If changes are made to the instance inserted into the circuit it will no longer be reflected in the gate in the circuit. This change was made to fix an issue when inserting a single parameterized Gate object into multiple circuits.

  • The function qiskit.result.marginal_counts() now, by default, does not modify the qiskit.result.Result instance parameter. Previously, the Result object was always modified in place. A new kwarg inplace has been added marginal_counts() which enables using the previous behavior when inplace=True is set.

  • The U3Gate definition has been changed to be in terms of the UGate class. The UGate class has no definition. It is therefore not possible to unroll every circuit in terms of U3 and CX anymore. Instead, U and CX can be used for every circuit.

  • The deprecated support for running Qiskit Terra with Python 3.5 has been removed. To use Qiskit Terra from this release onward you will now need to use at least Python 3.6. If you are using Python 3.5 the last version which will work is Qiskit Terra 0.15.2.

  • In the PulseBackendConfiguration in the hamiltonian attributes the vars field is now returned in a unit of Hz instead of the previously used GHz. This change was made to be consistent with the units used with the other attributes in the class.

  • The previously deprecated support for passing in a dictionary as the first positional argument to DAGNode constructor has been removed. Using a dictonary for the first positional argument was deprecated in the 0.13.0 release. To create a DAGNode object now you should directly pass the attributes as kwargs on the constructor.

  • The keyword arguments for the circuit gate methods (for example: qiskit.circuit.QuantumCircuit.cx) q, ctl*, and tgt*, which were deprecated in the 0.12.0 release, have been removed. Instead, only qubit, control_qubit* and target_qubit* can be used as named arguments for these methods.

  • The previously deprecated module qiskit.extensions.standard has been removed. This module has been deprecated since the 0.14.0 release. The qiskit.circuit.library can be used instead. Additionally, all the gate classes previously in qiskit.extensions.standard are still importable from qiskit.extensions.

  • The previously deprecated gates in the module qiskit.extensions.quantum_initializer: DiagGate, UCG`, UCPauliRotGate, UCRot, UCRXGate, UCX, UCRYGate, UCY, UCRZGate, UCZ have been removed. These were all deprecated in the 0.14.0 release and have alternatives available in the circuit library (qiskit.circuit.library).

  • The previously deprecated qiskit.circuit.QuantumCircuit gate method iden() has been removed. This was deprecated in the 0.13.0 release and i() or id() can be used instead.

Deprecation Notes

  • The use of a numpy.ndarray for a parameter in the params kwarg for the constructor of the Gate class and subclasses has been deprecated and will be removed in future releases. This was done as part of the refactoring of how parms type checking is handled for the Gate class. If you have a custom gate class which is a subclass of Gate directly (or via a different parent in the hierarchy) that accepts an ndarray parameter, you should define a custom validate_parameter() method for your class that will return the allowed parameter type. For example:

    def validate_parameter(self, parameter):
    """Custom gate parameter has to be an ndarray."""
    if isinstance(parameter, numpy.ndarray):
        return parameter
    else:
        raise CircuitError("invalid param type {0} in gate "
                           "{1}".format(type(parameter), self.name))

  • The num_ancilla_qubits property of the PiecewiseLinearPauliRotations and PolynomialPauliRotations classes has been deprecated and will be removed in a future release. Instead the property num_ancillas should be used instead. This was done to make it consistent with the QuantumCircuit method num_ancillas().

  • The qiskit.circuit.library.MSGate class has been deprecated, but will remain in place to allow loading of old jobs. It has been replaced with the qiskit.circuit.library.GMS class which should be used instead.

  • The MSBasisDecomposer transpiler pass has been deprecated and will be removed in a future release. The qiskit.transpiler.passes.BasisTranslator pass can be used instead.

  • The QuantumCircuit methods u1, u2 and u3 are now deprecated. Instead the following replacements can be used.

    u1(theta) = p(theta) = u(0, 0, theta)
u2(phi, lam) = u(pi/2, phi, lam) = p(pi/2 + phi) sx p(pi/2 lam)
u3(theta, phi, lam) = u(theta, phi, lam) = p(phi + pi) sx p(theta + pi) sx p(lam)

    The gate classes themselves, U1Gate, U2Gate and U3Gate remain, to allow loading of old jobs.

Bug Fixes

  • The Result class’s methods data(), get_memory(), get_counts(), get_unitary(), and get_statevector ` will now emit a warning when the ``experiment`() kwarg is specified for attempting to fetch results using either a QuantumCircuit or Schedule instance, when more than one entry matching the instance name is present in the Result object. Note that only the first entry matching this name will be returned. Fixes #3207(opens in a new tab)
  • The qiskit.circuit.QuantumCircuit method append() can now be used to insert one parameterized gate instance into multiple circuits. This fixes a previous issue where inserting a single parameterized Gate object into multiple circuits would cause failures when one circuit had a parameter assigned. Fixes #4697(opens in a new tab)
  • Previously the qiskit.execute.execute() function would incorrectly disallow both the backend and pass_manager kwargs to be specified at the same time. This has been fixed so that both backend and pass_manager can be used together on calls to execute(). Fixes #5037(opens in a new tab)
  • The QuantumCircuit method unitary() method has been fixed to accept a single integer for the qarg argument (when adding a 1-qubit unitary). The allowed types for the qargs argument are now int, Qubit, or a list of integers. Fixes #4944(opens in a new tab)
  • Previously, calling inverse() on a BlueprintCircuit object could fail if its internal data property was not yet populated. This has been fixed so that the calling inverse() will populate the internal data before generating the inverse of the circuit. Fixes #5140(opens in a new tab)
  • Fixed an issue when creating a qiskit.result.Counts object from an empty data dictionary. Now this will create an empty Counts object. The most_frequent() method is also updated to raise a more descriptive exception when the object is empty. Fixes #5017(opens in a new tab)
  • Fixes a bug where setting ctrl_state of a UnitaryGate would be applied twice; once in the creation of the matrix for the controlled unitary and again when calling the definition() method of the qiskit.circuit.ControlledGate class. This would give the appearence that setting ctrl_state had no effect.
  • Previously the ControlledGate method inverse() would not preserve the ctrl_state parameter in some cases. This has been fixed so that calling inverse() will preserve the value ctrl_state in its output.
  • Fixed a bug in the mpl output backend of the circuit drawer qiskit.circuit.QuantumCircuit.draw() and qiskit.visualization.circuit_drawer() that would cause the drawer to fail if the style kwarg was set to a string. The correct behavior would be to treat that string as a path to a JSON file containing the style sheet for the visualization. This has been fixed, and warnings are raised if the JSON file for the style sheet can’t be loaded.
  • Fixed an error where loading a QASM file via from_qasm_file() or from_qasm_str() would fail if a u, phase(p), sx, or sxdg gate were present in the QASM file. Fixes #5156(opens in a new tab)
  • Fixed a bug that would potentially cause registers to be mismapped when unrolling/decomposing a gate defined with only one 2-qubit operation.

Aer 0.7.0

Prelude

This 0.7.0 release includes numerous performance improvements and significant enhancements to the simulator interface, and drops support for Python 3.5. The main interface changes are configurable simulator backends, and constructing preconfigured simulators from IBMQ backends. Noise model an basis gate support has also been extended for most of the Qiskit circuit library standard gates, including new support for 1 and 2-qubit rotation gates. Performance improvements include adding SIMD support to the density matrix and unitary simulation methods, reducing the used memory and improving the performance of circuits using statevector and density matrix snapshots, and adding support for Kraus instructions to the gate fusion circuit optimization for greatly improving the performance of noisy statevector simulations.

New Features

  • Adds basis gate support for the qiskit.circuit.Delay instruction to the StatevectorSimulator, UnitarySimulator, and QasmSimulator. Note that this gate is treated as an identity gate during simulation and the delay length parameter is ignored.

  • Adds basis gate support for the single-qubit gate qiskit.circuit.library.UGate to the StatevectorSimulator, UnitarySimulator, and the "statevector", "density_matrix", "matrix_product_state", and "extended_stabilizer" methods of the QasmSimulator.

  • Adds basis gate support for the phase gate qiskit.circuit.library.PhaseGate to the StatevectorSimulator, StatevectorSimulator, UnitarySimulator, and the "statevector", "density_matrix", "matrix_product_state", and "extended_stabilizer" methods of the QasmSimulator.

  • Adds basis gate support for the controlled-phase gate qiskit.circuit.library.CPhaseGate to the StatevectorSimulator, StatevectorSimulator, UnitarySimulator, and the "statevector", "density_matrix", and "matrix_product_state" methods of the QasmSimulator.

  • Adds support for the multi-controlled phase gate qiskit.circuit.library.MCPhaseGate to the StatevectorSimulator, UnitarySimulator, and the "statevector" method of the QasmSimulator.

  • Adds support for the (X)\sqrt(X) gate qiskit.circuit.library.SXGate to the StatevectorSimulator, UnitarySimulator, and QasmSimulator.

  • Adds support for 1 and 2-qubit Qiskit circuit library rotation gates RXGate, RYGate, RZGate, RGate, RXXGate, RYYGate, RZZGate, RZXGate to the StatevectorSimulator, UnitarySimulator, and the "statevector" and "density_matrix" methods of the QasmSimulator.

  • Adds support for multi-controlled rotation gates "mcr", "mcrx", "mcry", "mcrz" to the StatevectorSimulator, UnitarySimulator, and the "statevector" method of the QasmSimulator.

  • Make simulator backends configurable. This allows setting persistant options such as simulation method and noise model for each simulator backend object.

    The QasmSimulator and PulseSimulator can also be configured from an IBMQBackend backend object using the :meth:`~qiskit.providers.aer.QasmSimulator.from_backend method. For the QasmSimulator this will configure the coupling map, basis gates, and basic device noise model based on the backend configuration and properties. For the PulseSimulator the system model and defaults will be configured automatically from the backend configuration, properties and defaults.

    For example a noisy density matrix simulator backend can be constructed as QasmSimulator(method='density_matrix', noise_model=noise_model), or an ideal matrix product state simulator as QasmSimulator(method='matrix_product_state').

    A benefit is that a PulseSimulator instance configured from a backend better serves as a drop-in replacement to the original backend, making it easier to swap in and out a simulator and real backend, e.g. when testing code on a simulator before using a real backend. For example, in the following code-block, the PulseSimulator is instantiated from the FakeArmonk() backend. All configuration and default data is copied into the simulator instance, and so when it is passed as an argument to assemble, it behaves as if the original backend was supplied (e.g. defaults from FakeArmonk will be present and used by assemble).

    armonk_sim = qiskit.providers.aer.PulseSimulator.from_backend(FakeArmonk())
pulse_qobj = assemble(schedules, backend=armonk_sim)
armonk_sim.run(pulse_qobj)

    While the above example is small, the demonstrated ‘drop-in replacement’ behavior should greatly improve the usability in more complicated work-flows, e.g. when calibration experiments are constructed using backend attributes.

  • Adds support for qobj global phase to the StatevectorSimulator, UnitarySimulator, and statevector methods of the QasmSimulator.

  • Improves general noisy statevector simulation performance by adding a Kraus method to the gate fusion circuit optimization that allows applying gate fusion to noisy statevector simulations with general Kraus noise.

  • Use move semantics for statevector and density matrix snapshots for the “statevector” and “density_matrix” methods of the QasmSimulator if they are the final instruction in a circuit. This reduces the memory usage of the simulator improves the performance by avoiding copying a large array in the results.

  • Adds support for general Kraus QauntumError gate errors in the NoiseModel to the "matrix_product_state" method of the QasmSimulator.

  • Adds support for density matrix snapshot instruction qiskit.providers.aer.extensions.SnapshotDensityMatrix to the "matrix_product_state" method of the QasmSimulator.

  • Extends the SIMD vectorization of the statevector simulation method to the unitary matrix, superoperator matrix, and density matrix simulation methods. This gives roughtly a 2x performance increase general simulation using the UnitarySimulator, the "density_matrix" method of the QasmSimulator, gate fusion, and noise simulation.

  • Adds a custom vector class to C++ code that has better integration with Pybind11. This haves the memory requirement of the StatevectorSimulator by avoiding an memory copy during Python binding of the final simulator state.

Upgrade Notes

  • AER now uses Lapack to perform some matrix related computations. It uses the Lapack library bundled with OpenBlas (already available in Linux and Macos typical OpenBlas dsitributions; Windows version distributed with AER) or with the accelerate framework in MacOS.

  • The deprecated support for running qiskit-aer with Python 3.5 has been removed. To use qiskit-aer >=0.7.0 you will now need at least Python 3.6. If you are using Python 3.5 the last version which will work is qiskit-aer 0.6.x.

  • Updates gate fusion default thresholds so that gate fusion will be applied to circuits with of more than 14 qubits for statevector simulations on the StatevectorSimulator and QasmSimulator.

    For the "density_matrix" method of the QasmSimulator and for the UnitarySimulator gate fusion will be applied to circuits with more than 7 qubits.

    Custom qubit threshold values can be set using the fusion_threshold backend option ie backend.set_options(fusion_threshold=10)

  • Changes fusion_threshold backend option to apply fusion when the number of qubits is above the threshold, not equal or above the threshold, to match the behavior of the OpenMP qubit threshold parameter.

Deprecation Notes

  • qiskit.providers.aer.noise.NoiseModel.set_x90_single_qubit_gates() has been deprecated as unrolling to custom basis gates has been added to the qiskit transpiler. The correct way to use an X90 based noise model is to define noise on the Sqrt(X) "sx" or "rx" gate and one of the single-qubit phase gates "u1", "rx", or "p" in the noise model.
  • The variance kwarg of Snapshot instructions has been deprecated. This function computed the sample variance in the snapshot due to noise model sampling, not the variance due to measurement statistics so was often being used incorrectly. If noise modeling variance is required single shot snapshots should be used so variance can be computed manually in post-processing.

Bug Fixes

  • Fixes bug in the StatevectorSimulator that caused it to always run as CPU with double-precision without SIMD/AVX2 support even on systems with AVX2, or when single-precision or the GPU method was specified in the backend options.
  • Fixes some for-loops in C++ code that were iterating over copies rather than references of container elements.
  • Fixes a bug where snapshot data was always copied from C++ to Python rather than moved where possible. This will halve memory usage and improve simulation time when using large statevector or density matrix snapshots.
  • Fix State::snapshot_pauli_expval to return correct Y expectation value in stabilizer simulator. Refer to #895 <https://github.com/Qiskit/qiskit-aer/issues/895(opens in a new tab)> for more details.
  • The controller_execute wrappers have been adjusted to be functors (objects) rather than free functions. Among other things, this allows them to be used in multiprocessing.pool.map calls.
  • Add missing available memory checks for the StatevectorSimulator and UnitarySimulator. This throws an exception if the memory required to simulate the number of qubits in a circuit exceeds the available memory of the system.

Ignis 0.5.0

Prelude

This release includes a new module for expectation value measurement error mitigation, improved plotting functionality for quantum volume experiments, several bug fixes, and drops support for Python 3.5.

New Features

  • The qiskit.ignis.verification.randomized_benchmarking.randomized_benchmarking_seq() function allows an optional input of gate objects as interleaved_elem. In addition, the CNOT-Dihedral class qiskit.ignis.verification.randomized_benchmarking.CNOTDihedral has a new method to_instruction, and the existing from_circuit method has an optional input of an Instruction (in addition to QuantumCircuit).

  • The qiskit.ignis.verification.randomized_benchmarking.CNOTDihedral now contains the following new features. Initialization from various types of objects: CNOTDihedral, ScalarOp, QuantumCircuit, Instruction and Pauli. Converting to a matrix using to_matrix and to an operator using to_operator. Tensor product methods tensor and expand. Calculation of the adjoint, conjugate and transpose using conjugate, adjoint and transpose methods. Verify that an element is CNOTDihedral using is_cnotdihedral method. Decomposition method to_circuit of a CNOTDihedral element into a circuit was extended to allow any number of qubits, based on the function decompose_cnotdihedral_general.

  • Adds expectation value measurement error mitigation to the mitigation module. This supports using complete N-qubit assignment matrix, single-qubit tensored assignment matrix, or continuous time Markov process (CTMP) [1] measurement error mitigation when computing expectation values of diagonal operators from counts dictionaries. Expectation values are computed using the using the qiskit.ignis.mitigation.expectation_value() function.

    Calibration circuits for calibrating a measurement error mitigator are generated using the qiskit.ignis.mitigation.expval_meas_mitigator_circuits() function, and the result fitted using the qiskit.ignis.mitigation.ExpvalMeasMitigatorFitter class. The fitter returns a mitigator object can the be supplied as an argument to the expectation_value() function to apply mitigation.

    [1] S Bravyi, S Sheldon, A Kandala, DC Mckay, JM Gambetta,

    Mitigating measurement errors in multi-qubit experiments, arXiv:2006.14044 [quant-ph].

    Example:

    The following example shows calibrating a 5-qubit expectation value measurement error mitigator using the 'tensored' method.

    from qiskit import execute
from qiskit.test.mock import FakeVigo
import qiskit.ignis.mitigation as mit
 
backend = FakeVigo()
num_qubits = backend.configuration().num_qubits
 
# Generate calibration circuits
circuits, metadata = mit.expval_meas_mitigator_circuits(
    num_qubits, method='tensored')
result = execute(circuits, backend, shots=8192).result()
 
# Fit mitigator
mitigator = mit.ExpvalMeasMitigatorFitter(result, metadata).fit()
 
# Plot fitted N-qubit assignment matrix
mitigator.plot_assignment_matrix()

    The following shows how to use the above mitigator to apply measurement error mitigation to expectation value computations

    from qiskit import QuantumCircuit
 
# Test Circuit with expectation value -1.
qc = QuantumCircuit(num_qubits)
qc.x(range(num_qubits))
qc.measure_all()
 
# Execute
shots = 8192
seed_simulator = 1999
result = execute(qc, backend, shots=8192, seed_simulator=1999).result()
counts = result.get_counts(0)
 
# Expectation value of Z^N without mitigation
expval_nomit, error_nomit = mit.expectation_value(counts)
print('Expval (no mitigation): {:.2f} \u00B1 {:.2f}'.format(
    expval_nomit, error_nomit))
 
# Expectation value of Z^N with mitigation
expval_mit, error_mit = mit.expectation_value(counts,
    meas_mitigator=mitigator)
print('Expval (with mitigation): {:.2f} \u00B1 {:.2f}'.format(
    expval_mit, error_mit))

  • Adds Numba as an optional dependency. Numba is used to significantly increase the performance of the qiskit.ignis.mitigation.CTMPExpvalMeasMitigator class used for expectation value measurement error mitigation with the CTMP method.

  • Add two methods to qiskit.ignis.verification.quantum_volume.QVFitter.

    • qiskit.ignis.verification.quantum_volume.QVFitter.calc_z_value() to calculate z value in standard normal distribution using mean and standard deviation sigma. If sigma = 0, it raises a warning and assigns a small value (1e-10) for sigma so that the code still runs.
    • qiskit.ignis.verification.quantum_volume.QVFitter.calc_confidence_level() to calculate confidence level using z value.

  • Store confidence level even when hmean < 2/3 in qiskit.ignis.verification.quantum_volume.QVFitter.qv_success().

  • Add explanations for how to calculate statistics based on binomial distribution in qiskit.ignis.verification.quantum_volume.QVFitter.calc_statistics().

  • The qiskit.ignis.verification.QVFitter method plot_qv_data() has been updated to return a matplotlib.Figure object. Previously, it would not return anything. By returning a figure this makes it easier to integrate the visualizations into a larger matplotlib workflow.

  • The error bars in the figure produced by the qiskit.ignis.verification.QVFitter method qiskit.ignis.verification.QVFitter.plot_qv_data() has been updated to represent two-sigma confidence intervals. Previously, the error bars represent one-sigma confidence intervals. The success criteria of Quantum Volume benchmarking requires heavy output probability > 2/3 with one-sided two-sigma confidence (~97.7%). Changing error bars to represent two-sigma confidence intervals allows easily identification of success in the figure.

  • A new kwarg, figsize has been added to the qiskit.ignis.verification.QVFitter method qiskit.ignis.verification.QVFitter.plot_qv_data(). This kwarg takes in a tuple of the form (x, y) where x and y are the dimension in inches to make the generated plot.

  • The qiskit.ignis.verification.quantum_volume.QVFitter.plot_hop_accumulative() method has been added to plot heavy output probability (HOP) vs number of trials similar to Figure 2a of Quantum Volume 64 paper (arXiv:2008.08571(opens in a new tab)). HOP of individual trials are plotted as scatters and cummulative HOP are plotted in red line. Two-sigma confidence intervals are plotted as shaded area and 2/3 success threshold is plotted as dashed line.

  • The qiskit.ignis.verification.quantum_volume.QVFitter.plot_qv_trial() method has been added to plot individual trials, leveraging on the qiskit.visualization.plot_histogram() method from Qiskit Terra. Bitstring counts are plotted as overlapping histograms for ideal (hollow) and experimental (filled) values. Experimental heavy output probability are shown on the legend. Median probability is plotted as red dashed line.

Upgrade Notes

  • The deprecated support for running qiskit-ignis with Python 3.5 has been removed. To use qiskit-ignis >=0.5.0 you will now need at least Python 3.6. If you are using Python 3.5 the last version which will work is qiskit-ignis 0.4.x.

Bug Fixes

  • Fixing a bug in the class qiskit.ignis.verification.randomized_benchmarking.CNOTDihedral for elements with more than 5 quits.
  • Fix the confidence level threshold for qiskit.ignis.verification.quantum_volume.QVFitter.qv_success() to 0.977 corresponding to z = 2 as defined by the QV paper Algorithm 1.
  • Fix a bug at qiskit.ignis.verification.randomized_benchmarking.randomized_benchmarking_seq() which caused all the subsystems with the same size in the given rb_pattern to have the same gates when a ‘rand_seed’ parameter was given to the function.

Aqua 0.8.0

Prelude

This release introduces an interface for running the available methods for Bosonic problems. In particular we introduced a full interface for running vibronic structure calculations.

This release introduces an interface for excited states calculations. It is now easier for the user to create a general excited states calculation. This calculation is based on a Driver which provides the relevant information about the molecule, a Transformation which provides the information about the mapping of the problem into a qubit Hamiltonian, and finally a Solver. The Solver is the specific way which the excited states calculation is done (the algorithm). This structure follows the one of the ground state calculations. The results are modified to take lists of expectation values instead of a single one. The QEOM and NumpyEigensolver are adapted to the new structure. A factory is introduced to run a numpy eigensolver with a specific filter (to target states of specific symmetries).

VQE expectation computation with Aer qasm_simulator now defaults to a computation that has the expected shot noise behavior.

New Features

  • Introduced an option warm_start that should be used when tuning other options does not help. When this option is enabled, a relaxed problem (all variables are continuous) is solved first and the solution is used to initialize the state of the optimizer before it starts the iterative process in the solve method.

  • The amplitude estimation algorithms now use QuantumCircuit objects as inputs to specify the A- and Q operators. This change goes along with the introduction of the GroverOperator in the circuit library, which allows an intuitive and fast construction of different Q operators. For example, a Bernoulli-experiment can now be constructed as

    import numpy as np
from qiskit import QuantumCircuit
from qiskit.aqua.algorithms import AmplitudeEstimation
 
probability = 0.5
angle = 2 * np.sqrt(np.arcsin(probability))
a_operator = QuantumCircuit(1)
a_operator.ry(angle, 0)
 
# construct directly
q_operator = QuantumCircuit(1)
q_operator.ry(2 * angle, 0)
 
# construct via Grover operator
from qiskit.circuit.library import GroverOperator
oracle = QuantumCircuit(1)
oracle.z(0)  # good state = the qubit is in state |1>
q_operator = GroverOperator(oracle, state_preparation=a_operator)
 
# use default construction in QAE
q_operator = None
 
ae = AmplitudeEstimation(a_operator, q_operator)

  • Add the possibility to compute Conditional Value at Risk (CVaR) expectation values.

    Given a diagonal observable H, often corresponding to the objective function of an optimization problem, we are often not as interested in minimizing the average energy of our observed measurements. In this context, we are satisfied if at least some of our measurements achieve low energy. (Note that this is emphatically not the case for chemistry problems).

    To this end, one might consider using the best observed sample as a cost function during variational optimization. The issue here, is that this can result in a non-smooth optimization surface. To resolve this issue, we can smooth the optimization surface by using not just the best observed sample, but instead average over some fraction of best observed samples. This is exactly what the CVaR estimator accomplishes [1].

    Let α\alpha be a real number in [0,1][0,1] which specifies the fraction of best observed samples which are used to compute the objective function. Observe that if α=1\alpha = 1, CVaR is equivalent to a standard expectation value. Similarly, if α=0\alpha = 0, then CVaR corresponds to using the best observed sample. Intermediate values of α\alpha interpolate between these two objective functions.

    The functionality to use CVaR is included into the operator flow through a new subclass of OperatorStateFn called CVaRMeasurement. This new StateFn object is instantied in the same way as an OperatorMeasurement with the exception that it also accepts an alpha parameter and that it automatically enforces the is_measurement attribute to be True. Observe that it is unclear what a CVaRStateFn would represent were it not a measurement.

    Examples:

    qc = QuantumCircuit(1)
qc.h(0)
op = CVaRMeasurement(Z, alpha=0.5) @ CircuitStateFn(primitive=qc, coeff=1.0)
result = op.eval()

    Similarly, an operator corresponding to a standard expectation value can be converted into a CVaR expectation using the CVaRExpectation converter.

    Examples:

    qc = QuantumCircuit(1)
qc.h(0)
op = ~StateFn(Z) @ CircuitStateFn(primitive=qc, coeff=1.0)
cvar_expecation = CVaRExpectation(alpha=0.1).convert(op)
result = cvar_expecation.eval()

    See [1] for additional details regarding this technique and it’s empircal performance.

    References:

    [1]: Barkoutsos, P. K., Nannicini, G., Robert, A., Tavernelli, I., and Woerner, S.,

    “Improving Variational Quantum Optimization using CVaR” arXiv:1907.04769(opens in a new tab)

  • New interface Eigensolver for Eigensolver algorithms.

  • An interface for excited states calculation has been added to the chemistry module. It is now easier for the user to create a general excited states calculation. This calculation is based on a Driver which provides the relevant information about the molecule, a Transformation which provides the information about the mapping of the problem into a qubit Hamiltonian, and finally a Solver. The Solver is the specific way which the excited states calculation is done (the algorithm). This structure follows the one of the ground state calculations. The results are modified to take lists of expectation values instead of a single one. The QEOM and NumpyEigensolver are adapted to the new structure. A factory is introduced to run a numpy eigensolver with a specific filter (to target states of specific symmetries).

  • In addition to the workflows for solving Fermionic problems, interfaces for calculating Bosonic ground and excited states have been added. In particular we introduced a full interface for running vibronic structure calculations.

  • The OrbitalOptimizationVQE has been added as new ground state solver in the chemistry module. This solver allows for the simulatneous optimization of the variational parameters and the orbitals of the molecule. The algorithm is introduced in Sokolov et al., The Journal of Chemical Physics 152 (12).

  • A new algorithm has been added: the Born Openheimer Potential Energy surface for the calculation of potential energy surface along different degrees of freedom of the molecule. The algorithm is called BOPESSampler. It further provides functionalities of fitting the potential energy surface to an analytic function of predefined potentials.

  • A feasibility check of the obtained solution has been added to all optimizers in the optimization stack. This has been implemented by adding two new methods to QuadraticProgram: * get_feasibility_info(self, x: Union[List[float], np.ndarray]) accepts an array and returns whether this solution is feasible and a list of violated variables(violated bounds) and a list of violated constraints. * is_feasible(self, x: Union[List[float], np.ndarray]) accepts an array and returns whether this solution is feasible or not.

  • Add circuit-based versions of FixedIncomeExpectedValue, EuropeanCallDelta, GaussianConditionalIndependenceModel and EuropeanCallExpectedValue to qiskit.finance.applications.

  • Gradient Framework. qiskit.operators.gradients Given an operator that represents either a quantum state resp. an expectation value, the gradient framework enables the evaluation of gradients, natural gradients, Hessians, as well as the Quantum Fisher Information.

    Suppose a parameterized quantum state |ψ(θ)〉 = V(θ)|ψ〉 with input state |ψ〉 and parametrized Ansatz V(θ), and an Operator O(ω).

    Gradients: We want to compute d⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dωd⟨ψ(θ)|O(ω)|ψ(θ)〉/ dω resp. d⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dθd⟨ψ(θ)|O(ω)|ψ(θ)〉/ dθ resp. d⟨ψ(θ)∣i〉⟨i∣ψ(θ)〉/dθd⟨ψ(θ)|i〉⟨i|ψ(θ)〉/ dθ.

    The last case corresponds to the gradient w.r.t. the sampling probabilities of |ψ(θ). These gradients can be computed with different methods, i.e. a parameter shift, a linear combination of unitaries and a finite difference method.

    Examples:

    x = Parameter('x')
ham = x * X
a = Parameter('a')
 
q = QuantumRegister(1)
qc = QuantumCircuit(q)
qc.h(q)
qc.p(params[0], q[0])
op = ~StateFn(ham) @ CircuitStateFn(primitive=qc, coeff=1.)
 
value_dict = {x: 0.1, a: np.pi / 4}
 
ham_grad = Gradient(grad_method='param_shift').convert(operator=op, params=[x])
ham_grad.assign_parameters(value_dict).eval()
 
state_grad = Gradient(grad_method='lin_comb').convert(operator=op, params=[a])
state_grad.assign_parameters(value_dict).eval()
 
prob_grad = Gradient(grad_method='fin_diff').convert(operator=CircuitStateFn(primitive=qc, coeff=1.),
                                                     params=[a])
prob_grad.assign_parameters(value_dict).eval()

    Hessians: We want to compute d2⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dω2d^2⟨ψ(θ)|O(ω)|ψ(θ)〉/ dω^2 resp. d2⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dθ2d^2⟨ψ(θ)|O(ω)|ψ(θ)〉/ dθ^2 resp. d2⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dθdωd^2⟨ψ(θ)|O(ω)|ψ(θ)〉/ dθdω resp. d2⟨ψ(θ)∣i〉⟨i∣ψ(θ)〉/dθ2d^2⟨ψ(θ)|i〉⟨i|ψ(θ)〉/ dθ^2.

    The last case corresponds to the Hessian w.r.t. the sampling probabilities of |ψ(θ). Just as the first order gradients, the Hessians can be evaluated with different methods, i.e. a parameter shift, a linear combination of unitaries and a finite difference method. Given a tuple of parameters Hessian().convert(op, param_tuple) returns the value for the second order derivative. If a list of parameters is given Hessian().convert(op, param_list) returns the full Hessian for all the given parameters according to the given parameter order.

    QFI: The Quantum Fisher Information QFI is a metric tensor which is representative for the representation capacity of a parameterized quantum state |ψ(θ)〉 = V(θ)|ψ〉 generated by an input state |ψ〉 and a parametrized Ansatz V(θ). The entries of the QFI for a pure state read [QFI]kl=Re[〈∂kψ∣∂lψ〉−〈∂kψ∣ψ〉〈ψ∣∂lψ〉]∗4[QFI]kl= Re[〈∂kψ|∂lψ〉−〈∂kψ|ψ〉〈ψ|∂lψ〉] * 4.

    Just as for the previous derivative types, the QFI can be computed using different methods: a full representation based on a linear combination of unitaries implementation, a block-diagonal and a diagonal representation based on an overlap method.

    Examples:

    q = QuantumRegister(1)
qc = QuantumCircuit(q)
qc.h(q)
qc.p(params[0], q[0])
op = ~StateFn(ham) @ CircuitStateFn(primitive=qc, coeff=1.)
 
value_dict = {x: 0.1, a: np.pi / 4}
qfi = QFI('lin_comb_full').convert(operator=CircuitStateFn(primitive=qc, coeff=1.), params=[a])
qfi.assign_parameters(value_dict).eval()

    The combination of the QFI and the gradient lead to a special form of a gradient, namely

    NaturalGradients: The natural gradient is a special gradient method which rescales a gradient w.r.t. a state parameter with the inverse of the corresponding Quantum Fisher Information (QFI) QFI−1d⟨ψ(θ)∣O(ω)∣ψ(θ)〉/dθQFI^-1 d⟨ψ(θ)|O(ω)|ψ(θ)〉/ dθ. Hereby, we can choose a gradient as well as a QFI method and a regularization method which is used together with a least square solver instead of exact invertion of the QFI:

    Examples:

    op = ~StateFn(ham) @ CircuitStateFn(primitive=qc, coeff=1.)
nat_grad = NaturalGradient(grad_method='lin_comb, qfi_method='lin_comb_full', \
                           regularization='ridge').convert(operator=op, params=params)

    The gradient framework is also compatible with the optimizers from qiskit.aqua.components.optimizers. The derivative classes come with a gradient_wrapper() function which returns the corresponding callable.

  • Introduces transformations for the fermionic and bosonic transformation of a problem instance. Transforms the fermionic operator to qubit operator. Respective class for the transformation is fermionic_transformation Introduces in algorithms ground_state_solvers for the calculation of ground state properties. The calculation can be done either using an MinimumEigensolver or using AdaptVQE Introduces chemistry/results where the eigenstate_result and the electronic_structure_result are also used for the algorithms. Introduces Minimum Eigensolver factories minimum_eigensolver_factories where chemistry specific minimum eigensolvers can be initialized Introduces orbital optimization vqe oovqe as a ground state solver for chemistry applications

  • New Algorithm result classes:

    Grover method _run() returns class GroverResult. AmplitudeEstimation method _run() returns class AmplitudeEstimationResult. IterativeAmplitudeEstimation method _run() returns class IterativeAmplitudeEstimationResult. MaximumLikelihoodAmplitudeEstimation method _run() returns class MaximumLikelihoodAmplitudeEstimationResult.

    All new result classes are backwards compatible with previous result dictionary.

  • New Linear Solver result classes:

    HHL method _run() returns class HHLResult. NumPyLSsolver method _run() returns class NumPyLSsolverResult.

    All new result classes are backwards compatible with previous result dictionary.

  • MinimumEigenOptimizationResult now exposes properties: samples and eigensolver_result. The latter is obtained from the underlying algorithm used by the optimizer and specific to the algorithm. RecursiveMinimumEigenOptimizer now returns an instance of the result class RecursiveMinimumEigenOptimizationResult which in turn may contains intermediate results obtained from the underlying algorithms. The dedicated result class exposes properties replacements and history that are specific to this optimizer. The depth of the history is managed by the history parameter of the optimizer.

  • GroverOptimizer now returns an instance of GroverOptimizationResult and this result class exposes properties operation_counts, n_input_qubits, and n_output_qubits directly. These properties are not available in the raw_results dictionary anymore.

  • SlsqpOptimizer now returns an instance of SlsqpOptimizationResult and this result class exposes additional properties specific to the SLSQP implementation.

  • Support passing QuantumCircuit objects as generator circuits into the QuantumGenerator.

  • Removes the restriction to real input vectors in CircuitStateFn.from_vector. The method calls extensions.Initialize. The latter explicitly supports (in API and documentation) complex input vectors. So this restriction seems unnecessary.

  • Simplified AbelianGrouper using a graph coloring algorithm of retworkx. It is faster than the numpy-based coloring algorithm.

  • Allow calling eval on state function objects with no argument, which returns the VectorStateFn representation of the state function. This is consistent behavior with OperatorBase.eval, which returns the MatrixOp representation, if no argument is passed.

  • Adds max_iterations to the VQEAdapt class in order to allow limiting the maximum number of iterations performed by the algorithm.

  • VQE expectation computation with Aer qasm_simulator now defaults to a computation that has the expected shot noise behavior. The special Aer snapshot based computation, that is much faster, with the ideal output similar to state vector simulator, may still be chosen but like before Aqua 0.7 it now no longer defaults to this but can be chosen.

Upgrade Notes

  • Extension of the previous Analytic Quantum Gradient Descent (AQGD) classical optimizer with the AQGD with Epochs. Now AQGD performs the gradient descent optimization with a momentum term, analytic gradients, and an added customized step length schedule for parametrized quantum gates. Gradients are computed “analytically” using the quantum circuit when evaluating the objective function.
  • The deprecated support for running qiskit-aqua with Python 3.5 has been removed. To use qiskit-aqua >=0.8.0 you will now need at least Python 3.6. If you are using Python 3.5 the last version which will work is qiskit-aqua 0.7.x.
  • Added retworkx as a new dependency.

Deprecation Notes

  • The i_objective argument of the amplitude estimation algorithms has been renamed to objective_qubits.

  • TransformationType

  • QubitMappingType

  • Deprecate the CircuitFactory and derived types. The CircuitFactory has been introduced as temporary class when the QuantumCircuit missed some features necessary for applications in Aqua. Now that the circuit has all required functionality, the circuit factory can be removed. The replacements are shown in the following table.

    Circuit factory class               | Replacement
------------------------------------+-----------------------------------------------
CircuitFactory                      | use QuantumCircuit
                                    |
UncertaintyModel                    | -
UnivariateDistribution              | -
MultivariateDistribution            | -
NormalDistribution                  | qiskit.circuit.library.NormalDistribution
MultivariateNormalDistribution      | qiskit.circuit.library.NormalDistribution
LogNormalDistribution               | qiskit.circuit.library.LogNormalDistribution
MultivariateLogNormalDistribution   | qiskit.circuit.library.LogNormalDistribution
UniformDistribution                 | qiskit.circuit.library.UniformDistribution
MultivariateUniformDistribution     | qiskit.circuit.library.UniformDistribution
UnivariateVariationalDistribution   | use parameterized QuantumCircuit
MultivariateVariationalDistribution | use parameterized QuantumCircuit
                                    |
UncertaintyProblem                  | -
UnivariateProblem                   | -
MultivariateProblem                 | -
UnivariatePiecewiseLinearObjective  | qiskit.circuit.library.LinearAmplitudeFunction

  • The ising convert classes qiskit.optimization.converters.QuadraticProgramToIsing and qiskit.optimization.converters.IsingToQuadraticProgram have been deprecated and will be removed in a future release. Instead the qiskit.optimization.QuadraticProgram methods to_ising() and from_ising() should be used instead.

  • Deprecate the WeightedSumOperator which has been ported to the circuit library as WeightedAdder in qiskit.circuit.library.

  • Core Hamiltonian class is deprecated in favor of the FermionicTransformation Chemistry Operator class is deprecated in favor of the tranformations minimum_eigen_solvers/vqe_adapt is also deprecated and moved as an implementation of the ground_state_solver interface applications/molecular_ground_state_energy is deprecated in favor of ground_state_solver

  • Optimizer.SupportLevel nested enum is replaced by OptimizerSupportLevel and Optimizer.SupportLevel was removed. Use, for example, OptimizerSupportLevel.required instead of Optimizer.SupportLevel.required.

  • Deprecate the UnivariateVariationalDistribution and MultivariateVariationalDistribution as input to the QuantumGenerator. Instead, plain QuantumCircuit objects can be used.

  • Ignored fast and use_nx options of AbelianGrouper.group_subops to be removed in the future release.

  • GSLS optimizer class deprecated __init__ parameter max_iter in favor of maxiter. SPSA optimizer class deprecated __init__ parameter max_trials in favor of maxiter. optimize_svm function deprecated max_iters parameter in favor of maxiter. ADMMParameters class deprecated __init__ parameter max_iter in favor of maxiter.

Bug Fixes

  • The UCCSD excitation list, comprising single and double excitations, was not being generated correctly when an active space was explicitly provided to UCSSD via the active_(un)occupied parameters.
  • For the amplitude estimation algorithms, we define the number of oracle queries as number of times the Q operator/Grover operator is applied. This includes the number of shots. That factor has been included in MLAE and IQAE but was missing in the ‘standard’ QAE.
  • Fix CircuitSampler.convert, so that the is_measurement property is propagated to converted StateFns.
  • Fix double calculation of coefficients in :meth`~qiskit.aqua.operators.VectorStateFn.to_circuit_op`.
  • Calling PauliTrotterEvolution.convert on an operator including a term that is a scalar multiple of the identity gave an incorrect circuit, one that ignored the scalar coefficient. This fix includes the effect of the coefficient in the global_phase property of the circuit.
  • Make ListOp.num_qubits check that all ops in list have the same num_qubits Previously, the number of qubits in the first operator in the ListOp was returned. With this change, an additional check is made that all other operators also have the same number of qubits.
  • Make PauliOp.exp_i() generate the correct matrix with the following changes. 1) There was previously an error in the phase of a factor of 2. 2) The global phase was ignored when converting the circuit to a matrix. We now use qiskit.quantum_info.Operator, which is generally useful for converting a circuit to a unitary matrix, when possible.
  • Fixes the cyclicity detection as reported buggy in https://github.com/Qiskit/qiskit-aqua/issues/1184(opens in a new tab).

IBM Q Provider 0.11.0

Upgrade Notes

  • The deprecated support for running qiskit-ibmq-provider with Python 3.5 has been removed. To use qiskit-ibmq-provider >=0.11.0 you will now need at least Python 3.6. If you are using Python 3.5 the last version which will work is qiskit-ibmq-provider 0.10.x.
  • Prior to this release, websockets 7.0 was used for Python 3.6. With this release, websockets 8.0 or above is required for all Python versions. The package requirements have been updated to reflect this.
Was this page helpful?
Report a bug or request content on GitHub.