StronglyEntanglingLayers with two qubits

wongwsvincent · May 13, 2021, 12:57am

Hello!

I am using StronglyEntanglingLayers to apply multiple layers of single qubit rotations and CNOT-entanglers. I realized that two CNOT gates are applied in each layer when I have only two wires, but I would like to apply it once only per layer.

Is there a reason why StronglyEntanglingLayers doesn’t handle scenarios with two wires like BasicEntanglerLayers? Many thanks!

Relevant pennylane code:

github.com

PennyLaneAI/pennylane/blob/master/pennylane/templates/layers/strongly_entangling.py#L114


        for l in range(self.n_layers):


            for i in range(len(self.wires)):
                qml.Rot(
                    self.parameters[0][l, i, 0],
                    self.parameters[0][l, i, 1],
                    self.parameters[0][l, i, 2],
                    wires=self.wires[i],
                )


            if len(self.wires) > 1:
                for i in range(len(self.wires)):
                    act_on = self.wires.subset([i, i + self.ranges[l]], periodic_boundary=True)
                    self.imprimitive(wires=act_on)


    return tape


@staticmethod
def shape(n_layers, n_wires):
    r"""Returns the expected shape of the weights tensor.

vs

github.com

PennyLaneAI/pennylane/blob/master/pennylane/templates/layers/basic_entangler.py#L156


    # first dimension of the weights tensor determines
    # the number of layers
    repeat = qml.math.shape(weights)[0]


    with qml.tape.QuantumTape() as tape:


        for layer in range(repeat):
            for i in range(len(self.wires)):
                self.rotation(weights[layer][i], wires=self.wires[i : i + 1])


            if len(self.wires) == 2:
                qml.CNOT(wires=self.wires)


            elif len(self.wires) > 2:
                for i in range(len(self.wires)):
                    w = self.wires.subset([i, i + 1], periodic_boundary=True)
                    qml.CNOT(wires=w)


    return tape


@staticmethod

Maria_Schuld · May 13, 2021, 2:17pm

Hey @wongwsvincent,

An excellent question

This is more a historical reason, BasicEntanglingLayers was coded up later when the issue of repeating the CNOT when we have 2 wires was not yet apparent to us.

I am in favour of changing the behaviour of StronglyEntanglingLayers! The best thing to do would be to open an issue where you explain what you would like to happen (also mention that this is already done in the other embedding). We can discuss in the issue how to go about changing it, since it will change people’s code. Either you or we can then make a PR!

wongwsvincent · May 13, 2021, 5:36pm

Hi @Maria_Schuld,

Thanks for the explanation! I have opened an issue as you suggested.

The implementation seems straightforward enough for me to contribute. I am happy to make a PR when there’s no strong objection to the idea

Maria_Schuld · May 14, 2021, 5:59am

Wonderful, thanks so much!

Topic		Replies	Views
Qnode with data-reupload and basic entangler layers PennyLane Help	2	886	July 14, 2021
QNN with varying layer WIDTH PennyLane Help	7	729	February 12, 2022
Drawing BasicEntanglerLayer circuit PennyLane Help	8	944	December 10, 2023
Building larger quantum circuits PennyLane Help	7	1663	January 27, 2020
Issues with simple QNN KerasLayer PennyLane Help	1	321	December 11, 2023

StronglyEntanglingLayers with two qubits

Related topics