Russian
Языки
English
Bengali
French
Hindi
Italian
Japanese
Korean
Malayalam
Russian
Spanish
Tamil
Turkish
Vietnamese
Shortcuts

Исходный код qiskit_machine_learning.algorithms.classifiers.pegasos_qsvc

# This code is part of Qiskit.
#
# (C) Copyright IBM 2022.
#
# This code is licensed under the Apache License, Version 2.0. You may
# obtain a copy of this license in the LICENSE.txt file in the root directory
# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
#
# Any modifications or derivative works of this code must retain this
# copyright notice, and modified files need to carry a notice indicating
# that they have been altered from the originals.

"""Pegasos Quantum Support Vector Classifier."""

import logging
from datetime import datetime
from typing import Optional, Dict

import numpy as np
from qiskit.utils import algorithm_globals
from sklearn.svm import SVC

from ...algorithms.serializable_model import SerializableModelMixin
from ...exceptions import QiskitMachineLearningError
from ...kernels.quantum_kernel import QuantumKernel

logger = logging.getLogger(__name__)


[документация]class PegasosQSVC(SVC, SerializableModelMixin): """ This class implements Pegasos Quantum Support Vector Classifier algorithm developed in [1] and includes overridden methods ``fit`` and ``predict`` from the ``SVC`` super-class. This implementation is adapted to work with quantum kernels. **Example** .. code-block:: python quantum_kernel = QuantumKernel() pegasos_qsvc = PegasosQSVC(quantum_kernel=quantum_kernel) pegasos_qsvc.fit(sample_train, label_train) pegasos_qsvc.predict(sample_test) **References** [1]: Shalev-Shwartz et al., Pegasos: Primal Estimated sub-GrAdient SOlver for SVM. `Pegasos for SVM <https://home.ttic.edu/~nati/Publications/PegasosMPB.pdf>`_ """ FITTED = 0 UNFITTED = 1 def __init__( self, quantum_kernel: Optional[QuantumKernel] = None, C: float = 1.0, num_steps: int = 1000, precomputed: bool = False, seed: Optional[int] = None, ) -> None: """ Args: quantum_kernel: QuantumKernel to be used for classification. Has to be ``None`` when a precomputed kernel is used. C: Positive regularization parameter. The strength of the regularization is inversely proportional to C. Smaller ``C`` induce smaller weights which generally helps preventing overfitting. However, due to the nature of this algorithm, some of the computation steps become trivial for larger ``C``. Thus, larger ``C`` improve the performance of the algorithm drastically. If the data is linearly separable in feature space, ``C`` should be chosen to be large. If the separation is not perfect, ``C`` should be chosen smaller to prevent overfitting. num_steps: number of steps in the Pegasos algorithm. There is no early stopping criterion. The algorithm iterates over all steps. precomputed: a boolean flag indicating whether a precomputed kernel is used. Set it to ``True`` in case of precomputed kernel. seed: a seed for the random number generator Raises: ValueError: - if ``quantum_kernel`` is passed and ``precomputed`` is set to ``True``. To use a precomputed kernel, ``quantum_kernel`` has to be of the ``None`` type. TypeError: - if ``quantum_instance`` neither instance of ``QuantumKernel`` nor ``None``. """ super().__init__(C=C) if precomputed: if quantum_kernel is not None: raise ValueError("'quantum_kernel' has to be None to use a precomputed kernel") else: if quantum_kernel is None: quantum_kernel = QuantumKernel() elif not isinstance(quantum_kernel, QuantumKernel): raise TypeError("'quantum_kernel' has to be of type None or QuantumKernel") self._quantum_kernel = quantum_kernel self._precomputed = precomputed self._num_steps = num_steps if seed is not None: algorithm_globals.random_seed = seed # these are the parameters being fit and are needed for prediction self._alphas: Optional[Dict[int, int]] = None self._x_train: Optional[np.ndarray] = None self._n_samples: Optional[int] = None self._y_train: Optional[np.ndarray] = None self._label_map: Optional[Dict[int, int]] = None self._label_pos: Optional[int] = None self._label_neg: Optional[int] = None # added to all kernel values to include an implicit bias to the hyperplane self._kernel_offset = 1 # for compatibility with the base SVC class. Set as unfitted. self.fit_status_ = PegasosQSVC.UNFITTED # pylint: disable=invalid-name
[документация] def fit( self, X: np.ndarray, y: np.ndarray, sample_weight: Optional[np.ndarray] = None ) -> "PegasosQSVC": """Fit the model according to the given training data. Args: X: Train features. For a callable kernel (an instance of ``QuantumKernel``) the shape should be ``(n_samples, n_features)``, for a precomputed kernel the shape should be ``(n_samples, n_samples)``. y: shape (n_samples), train labels . Must not contain more than two unique labels. sample_weight: this parameter is not supported, passing a value raises an error. Returns: ``self``, Fitted estimator. Raises: ValueError: - X and/or y have the wrong shape. - X and y have incompatible dimensions. - y includes more than two unique labels. - Pre-computed kernel matrix has the wrong shape and/or dimension. NotImplementedError: - when a sample_weight which is not None is passed. """ # check whether the data have the right format if np.ndim(X) != 2: raise ValueError("X has to be a 2D array") if np.ndim(y) != 1: raise ValueError("y has to be a 1D array") if len(np.unique(y)) != 2: raise ValueError("Only binary classification is supported") if X.shape[0] != y.shape[0]: raise ValueError("'X' and 'y' have to contain the same number of samples") if self._precomputed and X.shape[0] != X.shape[1]: raise ValueError( "For a precomputed kernel, X should be in shape (n_samples, n_samples)" ) if sample_weight is not None: raise NotImplementedError( "Parameter 'sample_weight' is not supported. All samples have to be weighed equally" ) # reset the fit state self.fit_status_ = PegasosQSVC.UNFITTED # the algorithm works with labels in {+1, -1} self._label_pos = np.unique(y)[0] self._label_neg = np.unique(y)[1] self._label_map = {self._label_pos: +1, self._label_neg: -1} # the training data are later needed for prediction self._x_train = X self._y_train = y self._n_samples = X.shape[0] # empty dictionary to represent sparse array self._alphas = {} t_0 = datetime.now() # training loop for step in range(1, self._num_steps + 1): # for every step, a random index (determining a random datum) is fixed i = algorithm_globals.random.integers(0, len(y)) value = self._compute_weighted_kernel_sum(i, X, training=True) if (self._label_map[y[i]] * self.C / step) * value < 1: # only way for a component of alpha to become non zero self._alphas[i] = self._alphas.get(i, 0) + 1 self.fit_status_ = PegasosQSVC.FITTED logger.debug("fit completed after %s", str(datetime.now() - t_0)[:-7]) return self
# pylint: disable=invalid-name
[документация] def predict(self, X: np.ndarray) -> np.ndarray: """ Perform classification on samples in X. Args: X: Features. For a callable kernel (an instance of ``QuantumKernel``) the shape should be ``(m_samples, n_features)``, for a precomputed kernel the shape should be ``(m_samples, n_samples)``. Where ``m`` denotes the set to be predicted and ``n`` the size of the training set. In that case, the kernel values in X have to be calculated with respect to the elements of the set to be predicted and the training set. Returns: An array of the shape (n_samples), the predicted class labels for samples in X. Raises: QiskitMachineLearningError: - predict is called before the model has been fit. ValueError: - Pre-computed kernel matrix has the wrong shape and/or dimension. """ if self.fit_status_ == PegasosQSVC.UNFITTED: raise QiskitMachineLearningError("The PegasosQSVC has to be fit first") if np.ndim(X) != 2: raise ValueError("X has to be a 2D array") if self._precomputed and self._n_samples != X.shape[1]: raise ValueError( "For a precomputed kernel, X should be in shape (m_samples, n_samples)" ) t_0 = datetime.now() y = np.zeros(X.shape[0]) for i in range(X.shape[0]): value = self._compute_weighted_kernel_sum(i, X, training=False) if value > 0: y[i] = self._label_pos else: y[i] = self._label_neg logger.debug("prediction completed after %s", str(datetime.now() - t_0)[:-7]) return y
def _compute_weighted_kernel_sum(self, index: int, X: np.ndarray, training: bool) -> float: """Helper function to compute the weighted sum over support vectors used for both training and prediction with the Pegasos algorithm. Args: index: fixed index distinguishing some datum X: Features training: flag indicating whether the loop is used within training or prediction Returns: Weighted sum of kernel evaluations employed in the Pegasos algorithm """ # non-zero indices corresponding to the support vectors support_indices = list(self._alphas.keys()) # for training if training: # support vectors x_supp = X[support_indices] # for prediction else: x_supp = self._x_train[support_indices] if not self._precomputed: # evaluate kernel function only for the fixed datum and the support vectors kernel = self._quantum_kernel.evaluate(X[index], x_supp) + self._kernel_offset else: kernel = X[index, support_indices] # map the training labels of the support vectors to {-1,1} y = np.array(list(map(self._label_map.get, self._y_train[support_indices]))) # weights for the support vectors alphas = np.array(list(self._alphas.values())) # this value corresponds to a sum of kernel values weighted by their labels and alphas value = np.sum(alphas * y * kernel) return value @property def quantum_kernel(self) -> QuantumKernel: """Returns quantum kernel""" return self._quantum_kernel @quantum_kernel.setter def quantum_kernel(self, quantum_kernel: QuantumKernel): """ Sets quantum kernel. If previously a precomputed kernel was set, it is reset to ``False``. """ self._quantum_kernel = quantum_kernel # quantum kernel is set, so we assume the kernel is not precomputed self._precomputed = False # reset training status self._reset_state() @property def num_steps(self) -> int: """Returns number of steps in the Pegasos algorithm.""" return self._num_steps @num_steps.setter def num_steps(self, num_steps: int): """Sets the number of steps to be used in the Pegasos algorithm.""" self._num_steps = num_steps # reset training status self._reset_state() @property def precomputed(self) -> bool: """Returns a boolean flag indicating whether a precomputed kernel is used.""" return self._precomputed @precomputed.setter def precomputed(self, precomputed: bool): """Sets the pre-computed kernel flag. If ``True`` is passed then the previous kernel is cleared. If ``False`` is passed then a new instance of ``QuantumKernel`` is created.""" self._precomputed = precomputed if precomputed: # remove the kernel, a precomputed will self._quantum_kernel = None else: # re-create a new default quantum kernel self._quantum_kernel = QuantumKernel() # reset training status self._reset_state() def _reset_state(self): """Resets internal data structures used in training.""" self.fit_status_ = PegasosQSVC.UNFITTED self._alphas = None self._x_train = None self._n_samples = None self._y_train = None self._label_map = None self._label_pos = None self._label_neg = None

© Copyright 2018, 2022, Qiskit Machine Learning Development Team.

Built with Sphinx using a theme provided by Read the Docs.