Using Strawberry Fields

Development

API

  1. Docs
  2. Module code
  3. strawberryfields
  4. strawberryfields.apps.train.embed

Source code for strawberryfields.apps.train.embed

# Copyright 2020 Xanadu Quantum Technologies Inc.

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at

#     http://www.apache.org/licenses/LICENSE-2.0

# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
r"""
Submodule for embedding trainable parameters into the GBS distribution.
"""

import numpy as np



[docs]class ExpFeatures:
    r"""Exponential embedding with feature vectors.

    Weights of the :math:`W` matrix in the :math:`WAW` parametrization are expressed as an exponential
    of the inner product between user-specified feature vectors and trainable parameters:
    :math:`w_i = \exp(-f^{(i)}\cdot\theta)`. The Jacobian, which encapsulates the derivatives of
    the weights with respect to the parameters can be computed straightforwardly as:
    :math:`\frac{d w_i}{d\theta_k} = -f^{(i)}_k w_i`.

    **Example usage:**

    >>> features = np.array([[0.1, 0.1, 0.1], [0.2, 0.2, 0.2], [0.3, 0.3, 0.3]])
    >>> embedding = ExpFeatures(features)
    >>> parameters = np.array([0.1, 0.2, 0.3])
    >>> embedding(parameters)
    [0.94176453 0.88692044 0.83527021]

    Args:
        features (np.array): Matrix of feature vectors where the i-th row is the i-th feature vector
    """

    def __init__(self, features):
        """Initializes the class for an input matrix whose rows are feature vectors"""
        self.features = features
        self.m, self.d = np.shape(features)

    def __call__(self, params):
        return self.weights(params)


[docs]    def weights(self, params):
        r"""Computes weights as a function of input parameters.

        Args:
            params (np.array): trainable parameters

        Returns:
            np.array: weights
        """
        if self.d != len(params):
            raise ValueError(
                "Dimension of parameter vector must be equal to dimension of feature vectors"
            )
        return np.exp(-self.features @ params)



[docs]    def jacobian(self, params):
        r"""Computes the Jacobian matrix of weights with respect to input parameters :math:`J_{
        ij} = \frac{\partial w_i}{\partial \theta_j}`.

        Args:
            params (np.array): trainable parameters

        Returns:
            np.array: Jacobian matrix of weights with respect to parameters
        """
        if self.d != len(params):
            raise ValueError(
                "Dimension of parameter vector must be equal to dimension of feature vectors"
            )
        w = self.weights(params)
        return -self.features * w[:, np.newaxis]




[docs]class Exp(ExpFeatures):
    r"""Simple exponential embedding.

    Weights of the :math:`W` matrix in the :math:`WAW` parametrization are expressed as an exponential
    of the trainable parameters: :math:`w_i = \exp(-\theta_i)`. The Jacobian, which encapsulates the
    derivatives of the weights with respect to the parameters can be computed straightforwardly as:
    :math:`\frac{d w_i}{d\theta_k} = -w_i\delta_{i,k}`.

    **Example usage:**

    >>> dim = 3
    >>> embedding = Exp(dim)
    >>> parameters = np.array([0.1, 0.2, 0.3])
    >>> embedding(parameters)
    [0.90483742 0.81873075 0.74081822]

    Args:
        dim (int): Dimension of the vector of parameters, which is equal to the number of modes
            in the GBS distribution
    """

    def __init__(self, dim):
        """The simple exponential mapping is a special case where the matrix of feature vectors
        is the identity"""
        super().__init__(np.eye(dim))