blueqat/blueqatSDK

✨ New cuQuantum Backend both on cuStateVec and cuTensorNet

• limited gate support with cusv and cutn backend.

✨ New default backend based on Tensor Network

• The back end has been changed to the tensor network. The previous backend environment can still be used with .run(backend="numpy").

✨ New QAOA function

• QAOA function is changed to the tensornetwork based.

📦 Get the single amplitude

• ```python
• Circuit(4).h[:].run(backend="quimb", amplitude="0101")
• ```

📦 Get the sample

• ```python
• Circuit(4).h[:].run(backend="quimb", shots=100)
• ```

📦 Get the expectation value of hamiltonian

• ```python
• from blueqat.pauli import Z
• hamiltonian = 1*Z[0]+1*Z[1]
• Circuit(4).x[:].run(backend="quimb", hamiltonian=hamiltonian)
• ```

📋 Changes

• Circuit drawing function with networkx

📋 Changes

• Correct `TGate.matrix()`, `TDagGate.matrix()`, `IGate.matrix()`
• New sampling feature is available in the numpy backend. This feature may effective for mid-circuit measurement.
• Circuit JSON scheme is updated due to experimental features.

📋 Changes

• Builtin-macros
• `mcx_with_ancilla(ctrl: list[int], target: int, ancilla: int) -> Circuit`
• `mcz_with_ancilla(ctrl: list[int], target: int, ancilla: int) -> Circuit`
• Remove wq module
• We recommend using dwave-ocean-sdk library for annealing task.
• `wq.pauli` function (It translates QUBO to pauli matrix for QAOA) is moved to `pauli.to_qubo` function.
• Internal implementation of Circuit
• It cheats type-checker likes pyright for suppressing errors.

📋 Changes

• Built-in macros.
• To activate, run `import blueqat.macros`.
• `draw`, `margolus`, `c3z`, `c4z`, `c3x`, `c4x`, `mcx_gray`, `mcz_gray`, `mcrx_gray`, `mcry_gray`, `mcrz_gray`, `mcr_gray`, `mcu_gray` are available.
• `blueqat.utils.circuit_to_unitary` is moved to `blueqat.circuit_utils.circuit_to_unitary`.

📋 Changes

• Add JSON serializer/deserializer. Refer [here](https://blueqat.readthedocs.io/en/latest/json.html).
• `Operation.create` method is added. All subclasses of `Operation` class must implement it.
• Remove circuit-local default backend, move `blueqat.circuit.GATE_SET` to `blueqat.gateset.GATE_SET`.
• Use `GATE_SET` directly is not recommended. Please use `blueqat.gateset.create` or `blueqat.gateset.get_op_type`.
• u1/u2/u3, cu1/cu2/cu3 gates are completely deleted.

📋 Changes

• Add the following methods. These methods are equivalent with `Circuit.run()` with `returns="..."` arguments, but these are typed.
• `Circuit.statevector() -> np.ndarray`
• `Circuit.shots(n: int) -> Counter`
• `Circuit.oneshot() -> Tuple[np.ndarray, str]`

📋 Changes

• Repair 1 qubit gate decomposition function
• Correct gamma factor of `U.matrix()`

📋 Changes

• Add `ryrz_decomposer`
• Debug wq (Ising model) module
• Debug phase/cphase gate matrix

📋 Changes

• U1(θ) → RZ(θ)
• U2(φ, λ) → U(π/2, φ, λ, -0.5 * (φ+λ))
• U3(θ, φ, λ) → U(θ, φ, λ, -0.5 * (φ+λ))
• CU1(θ) → CR(θ)
• CU2(φ, λ) → CU(π/2, φ, λ, -0.5 * (φ+λ))
• CU3(θ, φ, λ) → CU(θ, φ, λ, -0.5 * (φ+λ))

📋 Changes

• Input initial statevector for numpy and numba backend. #101
• Cache is not stored after running. If caching is required, use `circuit.run(save_cache=True)`.
• Correct qgate backend, SWAP gate error.

📋 Changes

• Make mat1 gate and basic transpiler. It will facilitate circuit optimizations and execution on the real quantum machine
• Add ZZ gate, this is for trapped-ion based machine's basis gate
• Update for next version of qgate (thank you, Morino san!)
• CCZ gate was not available. Now, it is available.

📋 Changes

• Add `reset` operation. This is equivalent with measurement without recording measured result and apply X gate when 1 is measured.
• Correct numba's global phase error.
• `opt` module is renamed to `wq`.

📋 Changes

• Improve speed for Pauli operator's `to_matrix()` method
• Add `n_qubits` property for Pauli operator term and expr.
• Correct pauli.Term.simplify method
• Solve the problem that constant value of `max_n` raises exception

📦 Random generated Hamiltonian

• Making CSC (sparse) matrix from random generated Hamiltonian.
• row: The number of qubits
• column: The number of terms
• cell: Time (sec.)

📦 Blueqat 0.3.16 (Previous release)

• 

📦 Blueqat 0.3.17 (This release)

• 

📦 Compare

• Time in v0.3.16 / Time in v0.3.17
• Greater than 1: v0.3.17 is faster than v0.3.16
• Less than 1: v 0.3.17 is slower than v0.3.16
• 
• For large Hamiltonian, this release is 100 times faster than previous release

📦 LiH Hamiltonian

• Compared with OpenFermion's `qubit_operator_sparse`.
• Normally, the number of qubits is 12, but can be add redundant qubit via n_qubits parameter.
• Measure 12 qubits to 19 qubits.
• 
• (*1): Due to lack of memory, so many swap out occurred. Blueqat implementation is less memory consumption than OpenFermion.
• New install
• `pip install blueqat`
• Update
• + 1 more

📦 `Vqe()` class

• When sampler is default, make sparse matrix and calculate expectation directly.
• This is for performance improvement.
• (If you're using default sampler, your VQE code is speed-up without any changes.)
• VQE is still under refactoring process and this API may be changed in future release, however, because this feature is very important, I want to merge this feature in the master.
• Many cases of VQEs, this change makes fantastic speed-ups.
• However, some cases, this change makes slow down.
• To make previous default behaviour, specify sampler as previous default sampler.
• ```py
• + 4 more

📦 `pauli.to_matrix()` qubit order is reordered.

• Previous: X[0] * Y[1] is X ⊗ Y
• New: X[0] * Y[1] is Y ⊗ X
• New behavior is equivalent with Blueqat circuit's statevector order.

📦 `Vqe()` class

• When sampler is default, make sparse matrix and calculate expectation directly.
• This is for performance improvement.
• (If you're using default sampler, your VQE code is speed-up without any changes.)
• VQE is still under refactoring process and this API may be changed in future release, however, because this feature is very important, I want to merge this feature in the master.
• Many cases of VQEs, this change makes fantastic speed-ups.
• However, some cases, this change makes slow down.
• To make previous default behaviour, specify sampler as previous default sampler.
• ```py
• + 4 more

✨ Features

• Add new gates, rxx, ryy, rzz, cswap.
• Add `Circuit.dagger`. `Circuit().rx(0.1)[0].cx[0, 1].t[0].dagger()` is `Circuit().tdg[0].cx[0, 1].rx(-0.1)[0]`.
• Add `mixer` argument for QaoaAnsatz. (This feature may be changed in the future.) This is for Quantum Alternative Operator Ansatz.
• Add `sparse` option for `pauli.Expr.to_unitary`. It returns scipy's sparse matrix. `(X[0] + Y[3]).to_matrix(sparse='csc') # csc matrix` The performance for larger qubits expr is better than (default's) dense matrix, but still not fast.

📦 Experimental

• Add `blueqat.experimental.utils.circuit_to_unitary`. It returns unitary matrix instead of statevector.
• Blueqat already has to_unitary method for sympy's unitary gate.
• However, this function runs the circuit with a backend which returns statevector.
• This feature is useful for debugging the circuit, and it's too heavy for larger circuit.

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

🐛 Bug Fix

• To avoiding qgate 0.2.1 bug, Blueqat had implementation of workaround.
• However, Blueqat workaround has bug.
• Now, bug fixed version qgate 0.2.2 was released.
• So, workaround was removed. It works well!

📦 Experimental

• Experimental feature is available in "blueqat.experimental" module.
• This is unstable features. It may modified or deleted.

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

📦 Circuit repr

• `Circuit.__repr__` method is implemented.
• Now, `Circuit` is printable.
• ```py
• c = Circuit().x[0].rz(1.2)[1]
• c.cx[0, 1]
• print(c)
• ```

📦 Adjust global phase

• These are Blueqat standard implementation of global phase.
• RZ and U1 gate:
• ```
• [e^-iθ/2, 0],
• [0, e^iθ/2]
• ```
• Phase gate (, S gate [θ=π/2], S† gate [θ=-π/2], T gate [θ=π/4], T† gate [θ=-π/4]):
• ```
• + 32 more

📦 numba backend

• Correct bugs in controlled gate of numba backend

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

📦 Enhance IBM Q backend "returns" option

• Add some `Circuit().run_with_ibmq(returns=...)` options:
• `shots`: (default) Returns measurement result
• `_exception`: If error occurs during importing qiskit, returns exception. otherwise, returns `None`.
• `draw`: (NEW): Draw the circuit via Qiskit QuantumCircuit.draw
• `qiskit_circuit`: (NEW): Returns Qiskit QuantumCircuit object
• `qiskit_job`: (NEW): Returns Qiskit Job object
• `qiskit_result`: (NEW) Returns Qiskit Result object
• Usage of draw option
• + 5 more

📦 Adjust global phase

• Adjust global phase for some gates.

✨ Add gates

• Add cy (controlled-y), ch (controlled-h), phase, cphase gates.
• phase gate is as same as rz gate except global phase.
• rz:
• [ e^-iθ/2, 0]
• [0, e^iθ/2]
• phase:
• [1, 0]
• [0, e^iθ]
• + 12 more

📦 numba backend

• Correct bugs in cnot gate and cz gate of numba backend

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

📦 `ignore_global` option for numpy and numba backend.

• Blueqat ignores global phase of statevectors.
• But it makes debugging difficult.
• We added `ignore_global` option to disable ignoring global phase.
• `Circuit().x[0].phase(0.123)[0].run()`
• => `array([0.+0.j, 1.+0.j])` Global phase is ignored.
• `Circuit().x[0].phase(0.123)[0].run(ignore_global=False)`
• => `array([0. +0.j , 0.99244503+0.12269009j])` Global phase is not ignored.
• Related to this change, we will modify global phases for some gates in future release.

✨ Modify IBMQ backend for new qiskit

• Modify for `qiskit~=0.11.1`

📦 Modify scipy requirements version

• Now, required scipy version is `scipy>1.1.0`

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

✨ New features

• Implemented Controlled-RX, RY, RZ.

📦 Install

• New install
• `pip install blueqat`
• Update
• `pip install -U blueqat`

✨ qgate backend (Thanks to Morino san): New backend for [qgate](https://github.com/shinmorino/qgate).

• qgate is required for this backend.
• Usage: `Circuit().h[:10].run(backend='qgate', shots=1000)` or `Circuit().h[:10].run(backend='qgate', runtime='cuda', shots=1000)`

✨ numba backend: New backend using numba.

• Usage: `Circuit().h[:10].run(backend='numba')`

✨ Add test for other backends

• To add backend, run `python -m pytest ./test --add-backend numba --add-backend qgate`.

🐛 Small modifications and debugging

• `pip install blueqat`