QuAIRKit 0.3.0 Release Note

Feature Rollout

• Probabilistic computation: internally supported parallel (pure & mixed state) computations for probabilistic outcomes generated by measurements or quasi-sampling:
  • Circuit.collapse: renamed as Circuit.measure. When desired_result is None, calling this attribute would store all mid-circuit measurement outcomes and the corresponding probability distribution.
  • Measure: same as above.
  • Circuit.locc: add one-way LOCC (unitary) protocol to the circuit. Detailed tutorials will be released in the next version.
  • State.index_select: replaced by State.prob_select, which can be called to selet the particular outcome.
• Controlled operation: QuAIRKit now support multi-control unitaries:
  • Circuit.control_oracle: can inputs multi-control indices and customized projections.
    For example, cir.control_oracle(u, [[0, 1], 2], P) is equivalent to inserting the unitary

$$(I - P) \otimes I + P \otimes U$$

• Quantum information functions: Advanced the quairkit.qinfo module with the following changes:
  • mutual_information: compute the the mutual information of a bipartite state
  • general_state_fidelity: compute the fidelity of two subnormalized states.
  • Nkron: now supports trivial inputs, and has an alias name nkron

Stability Update

• All matrix and channel representation functions are compatitble with NumPy inputs and outcomputs.
• Add code examples in docstrings of functions in database.
• Improve the efficiency and interfaces of PQCombNet, which is now imported by quairkit.application.
• Add bug fixes.

---

End of Release Notes

QuAIRKit 0.2.0 Release Note

Feature Rollout

• Qudit computation: QuAIRKit now support qudit systems in state evolution, circuit measurement, database generation and training. These covered functionals accept num_systems and system_dim as inputs.
  • argument num_systems is aliased with num_qubits
  • argument system_dim accepts both interger or a list of integers, representing the dimensions of all systems. Defaults to 2, i.e., all systems are qubits.
• Quantum information functions: Advanced the quairkit.qinfo module with the following changes:
  • partial_trace & partial_transpose: The interfaces of these two functions are aligned with PartialTrace and PartialTranspose in QETLAB. Moreover, these functions support qudit computations.
    • transpose_2 is now removed from quairkit.
    • partial_trace_discontiguous remains unchanged, but only supports qubit systems.
  • is_povm & is_pvm: Check whether the input set of matrices form a POVM or PVM.
  • mana_state & mana_channel: Compute the mana of input state/channel.
  • stab_nullity & stab_renyi: Compute the stabilizer nullity/Renyi entropy of input state
  • Almost all functions in quairkit.qinfo support batched array inputs.
• Training: Add training net PQCombNet for quantum circuit architectural search, can be imported by quairkit.ansatz.comb.
• State features: Add the following properties & methods for flexible usage
  • system_dim: User now can change the system dimensions of a State instance.
  • trace, transpose & permute: same as partial_trace, transpose & permute_systems, except there is no need to input data or specify system dimensions.
  • sqrt & log: same as sqrtm & logm

Stability Update

• Import regulation: Add __init__ to all top-level modules, so that the following commands

from quairkit.database import *
from quairkit.loss import *
from quairkit.qinfo import *

would not import some unexpected variables.

• GPU speedup: Significantly improve GPU performances on batch computations.
• Add quairkit.print_info function that can print all packages and hardware information for sanity check.
• Add bug fixes and API improvements.

Tutorials

• Rewrite 4 old tutorials to improve readability.
• Add 4 new introduction tutorials, including:
  • Hamiltonian in QuAIRKit
  • Quantum gates and quantum channels
  • Quantum information tools
  • Training parameterized quantum circuits
• Add 2 new feature tutorials, including:
  • Neural network setup customization
  • Introduction to qudit quantum computing

---

End of Release Notes

QuAIRKit 0.1.0 Release Note

QuAIRKit absorbs the features and semantics of Paddle Quantum v2.4.0, while making the following improvements:

Technical Improvements

• Backend Transition: transitioned the computational backend from Paddle Paddle to PyTorch, improving both stability and compatibility with existing packages.

• Modular Codebase: refactored the codebase into five discrete modules to streamline the user experience. This modular approach facilitates easier maintenance, scalability, and collaborative development.

• Batch Computation: internally supported batch computation for

  • quantum (noisy) circuit simulation;
  • expectations and measurement

Basic broadcasting rule in QuAIRKit:

Batched shape for operations	Batched shape for input states	Batched shape for output results
None	None	None
None	n	n
n	None	n
n	n	n
n	m	Error

• Pure & Mixed Auto-transition: there is no longer needed to specify the backend state_vector or density_matrix. QuAIRKit supports implicit backend transition between pure state and mixed state simulators.

Feature Rollout

• Random Seed Initialization: introduced quairkit.set_seed to initialize a random seed across all random functionals within the platform.

• Quantum computing data: Expanded the quairkit.database module with the following data:

  • phase & shift: phase and shift operators in quantum information.
  • universal_qudit: a parameterized universal qudit gate, based on parameterization with respect to Gell-Mann matrices.
  • Uf & Of: two quantum query models for Boolean functions.
  • A collection of set-based data, including:
    • std_basis & qft_basis: two quantum state bases for identity & QFT matrices, respectively.
    • bell_basis: a two-qubit Bell basis.
    • heisenberg_weyl: a set of Heisenberg-Weyl operators.
    • phase_space_point: a set of phase-space point operators in discrete phase space.
    • gell_mann: a set of Gell-Mann matrices.

• Quantum information functions: Expanded the quairkit.qinfo module with the following functions:

  • is_linear: ascertains the linearity of a given function.
  • create_matrix: constructs the matrix representation of a linear map between two vector spaces.
  • create_choi_repr: constructs the Choi representation of a linear map between two operator spaces.
  • trace_norm & p_norm: Schatten norm functions.
  • trace: Trace of input tensor along a specified axis.
  • hessian & gradient: The Hessian and general gradient of input functions.
  • decomp_1qubit & decomp_ctrl_1qubit: Decomposition of (controlled) qubit-unitary operations.

Resolved Issues

• Bug Fix: corrected various typos and inaccuracies previously present in the database.
• Controlled qubit-unitary: controlled qubit-unitary operations now require parameterization from 3 to 4 parameters.

Dependencies Update

• The minimum requirement for the Python environment has been changed to Python 3.8.
• Add pytest-xdist for unit tests.

Tutorials

Added four tutorials for feature introductions of QuAIRKit:

• circuit: quantum circuit & basic machine learning
• measure: quantum measurement
• state: quantum state
• batch: batch computation

---

End of Release Notes