HS.6.1d Incorrect: your cosine circuit isn't quite right yet!

Codebook
1

Hi everyone,

I’m working on the QSVT exercise where we need to implement the circuit_cos(t) and circuit_sin(t)

In my circuit_cos(t) implementation, I apply a Hadamard to the control wire, then call block_encode() once (controlled on wire 1), and finally apply all the phase gates from cos_projectors(t) in order, followed by another Hadamard.

However, I keep getting the error:

Incorrect: your cosine circuit isn't quite right yet!

Here is my entire code:

control_wires = [2,3]
work_wires = [4,5]

dev = qml.device("default.qubit")

@qml.qnode(dev)
def circuit_cos(t):
    """
    Uses QSVT to apply an approximate cosine transformation to a block-encoded matrix
    """

    qml.Hadamard(wires=1)  # QSP starts with Hadamard on the control wire

    qml.ctrl(block_encode, control=1)

    for gate in cos_projectors(t):
        qml.apply(gate)

    qml.Hadamard(wires=1)  

    return qml.state()

matrix_1 = qml.matrix(circuit_cos, wire_order=[1] + control_wires + work_wires)(0.5)
print("The cosine of the H*t for t = 0.5 is: \n")
print(np.round(matrix_1[: 2 ** len(work_wires), : 2 ** len(work_wires)], 4))

@qml.qnode(dev)
def circuit_sin(t):
    """
    Uses QSVT to apply an approximate sine transformation to a block-encoded matrix
    """

    qml.Hadamard(wires=1)  # QSP starts with Hadamard on the control wire

    qml.ctrl(block_encode, control=1)

    for gate in sin_projectors(t):
        qml.apply(gate)

    qml.Hadamard(wires=1)

    return qml.state()

matrix_2 = qml.matrix(circuit_sin, wire_order=[1] + control_wires + work_wires)(0.5)
print("The sine of the H*t for t = 0.5 is: \n")
print(np.round(matrix_2[: 2 ** len(work_wires), : 2 ** len(work_wires)], 4))

I’m not sure what I’m missing. Could it be the order of operations in the QSP sequence?

Any help or clarification would be greatly appreciated!

Thanks in advance!

2

Hi!
The problem statement says that you should apply QSVT using the cos and sin projectors.
Fist and last Hadamard are fine. However, the controlled operation should be a QSVT operation, not the block-encoded matrix. Use both documentations for qml.ctrl and qml.QSVT for guidance.
Do not forget to implement the adjoint part of QSVT as well as shown in the circuit figure in the theory section.
I hope this helps!