Release 0.12.0

New features since last release

• A new device, strawberryfields.tf, provides support for using Strawberry Fields TensorFlow backend from within PennyLane. (#50)

  dev = qml.device('strawberryfields.tf', wires=2, cutoff_dim=10)

  This device supports classical backpropagation when using the TensorFlow interface:

    @qml.qnode(dev, interface="tf", diff_method="backprop")
    def circuit(x, theta):
        qml.Displacement(x, 0, wires=0)
        qml.Beamsplitter(theta, 0, wires=[0, 1])
        return qml.probs(wires=0)

  Gradients will be computed using TensorFlow backpropagation:

  >>> x = tf.Variable(1.0)
  >>> theta = tf.Variable(0.543)
  >>> with tf.GradientTape() as tape:
  ...     res = circuit(x, theta)
  >>> jac = tape.jacobian(res, x)
  >>> print(jac)
  <tf.Tensor: shape=(1, 10), dtype=float32, numpy=
  array([[-7.0436597e-01,  1.8805575e-01,  3.2707882e-01,  1.4299491e-01,
           3.7763387e-02,  7.2306832e-03,  1.0900890e-03,  1.3535164e-04,
           1.3895189e-05,  9.9099987e-07]], dtype=float32)>

  For more details, please see the TF device documentation

• A new device, strawberryfields.gbs, provides support for training of the Gaussian boson sampling (GBS) distribution. (#47)

  dev = qml.device('strawberryfields.gbs', wires=4, cutoff_dim=4)

  This device allows the adjacency matrix A of a graph to be trained. The QNode must have a fixed structure:

  from pennylane_sf.ops import ParamGraphEmbed
  import numpy as np
  
  A = np.array([
      [0.0, 1.0, 1.0, 1.0],
      [1.0, 0.0, 1.0, 0.0],
      [1.0, 1.0, 0.0, 0.0],
      [1.0, 0.0, 0.0, 0.0]])
  n_mean = 2.5
  
  @qml.qnode(dev)
  def quantum_function(x):
      ParamGraphEmbed(x, A, n_mean, wires=range(4))
      return qml.probs(wires=range(4))

  Here, n_mean is the initial mean number of photons in the output GBS samples. The GBS probability distribution for a choice of trainable parameters x can then be accessed:

  >>> x = 0.9 * np.ones(4)
  >>> quantum_function(x)

  For more details, please see the gbs device documentation

Improvements

• Adds the ability for the StrawberryFieldsGBS device to use the reparametrization trick in sampling mode. (#55)

Bug fixes

• Applies minor fixes to RemoteEngine. (#53)

• Sets a fixed cutoff dimension for RemoteEngine. (#54)

• Adds unwrapping for operation parameters as indexing into NumPy arrays was added to PennyLane. (#56)

Contributors

This release contains contributions from (in alphabetical order):

Juan Miguel Arrazola, Thomas Bromley, Josh Izaac.