Python numpy.infty方法代码示例

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def extend(self, token, log_prob, state, decoder_output, encoder_mask, attn_dist, p_gen, coverage):
    """Return a NEW hypothesis, extended with the information from the latest step of beam search.

    Args:
      token: Integer. Latest token produced by beam search.
      log_prob: Float. Log prob of the latest token.
      state: Current decoder state, a LSTMStateTuple.
      attn_dist: Attention distribution from latest step. Numpy array shape (attn_length).
      p_gen: Generation probability on latest step. Float.
      coverage: Latest coverage vector. Numpy array shape (attn_length), or None if not using coverage.
    Returns:
      New Hypothesis for next step.
    """
    if FLAGS.avoid_trigrams and self._has_trigram(self.tokens + [token]):
        log_prob = -np.infty
    return Hypothesis(tokens = self.tokens + [token],
                      log_probs = self.log_probs + [log_prob],
                      state = state,
                      decoder_output= self.decoder_output + [decoder_output] if decoder_output is not None else [],
                      encoder_mask = self.encoder_mask + [encoder_mask] if encoder_mask is not None else [],
                      attn_dists = self.attn_dists + [attn_dist],
                      p_gens = self.p_gens + [p_gen],
                      coverage = coverage) 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def fit(self, X):
        """Fit GMM parameters to X

        Parameters
        ----------
        X : array-like, shape (n_samples, n_features)

        """
        # initialize
        self._initialize_parameters(X, self.random_state)
        lower_bound = -np.infty

        for n in range(self.n_iter):
            # E-step
            log_prob_norm, log_resp = self._e_step(X)

            # M-step
            self._m_step(X, log_resp)

            # check convergence
            back_lower_bound = lower_bound
            lower_bound = self._compute_lower_bound(
                log_resp, log_prob_norm) 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def __init__(self, reg_e=1., reg_cl=0.1,
                 max_iter=10, max_inner_iter=200, log=False,
                 tol=10e-9, verbose=False,
                 metric="sqeuclidean", norm=None,
                 distribution_estimation=distribution_estimation_uniform,
                 out_of_sample_map='ferradans', limit_max=np.infty):
        self.reg_e = reg_e
        self.reg_cl = reg_cl
        self.max_iter = max_iter
        self.max_inner_iter = max_inner_iter
        self.tol = tol
        self.log = log
        self.verbose = verbose
        self.metric = metric
        self.norm = norm
        self.distribution_estimation = distribution_estimation
        self.out_of_sample_map = out_of_sample_map
        self.limit_max = limit_max 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def test_monotonic_likelihood():
    # We check that each step of the each step of variational inference without
    # regularization improve monotonically the training set of the bound
    rng = np.random.RandomState(0)
    rand_data = RandomData(rng, scale=20)
    n_components = rand_data.n_components

    for prior_type in PRIOR_TYPE:
        for covar_type in COVARIANCE_TYPE:
            X = rand_data.X[covar_type]
            bgmm = BayesianGaussianMixture(
                weight_concentration_prior_type=prior_type,
                n_components=2 * n_components, covariance_type=covar_type,
                warm_start=True, max_iter=1, random_state=rng, tol=1e-4)
            current_lower_bound = -np.infty
            # Do one training iteration at a time so we can make sure that the
            # training log likelihood increases after each iteration.
            for _ in range(600):
                prev_lower_bound = current_lower_bound
                current_lower_bound = bgmm.fit(X).lower_bound_
                assert_greater_equal(current_lower_bound, prev_lower_bound)

                if bgmm.converged_:
                    break
            assert(bgmm.converged_) 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def __init__(self, data, max_mixture=10, threshold=0.1):
        """
        Class constructor, arguments include:
            data - data to build GMM model
            max_mixture - max number of Gaussian mixtures
            threshold - probability threhold to determine fense
        """
        self.data = data
        self.thresh = threshold
        lowest_bic = np.infty
        components = 1
        bic = []
        n_components_range = range(1, max_mixture + 1)
        for n_components in n_components_range:
            # Fit a Gaussian mixture with EM
            gmm = mixture.GaussianMixture(n_components=n_components,
                                          random_state=1005)
            gmm.fit(data)
            bic.append(gmm.bic(data))
            if bic[-1] < lowest_bic:
                lowest_bic = bic[-1]
                best_gmm = gmm
                components = n_components
        log.debug('best gmm components number: %d, bic %f ', components, lowest_bic)
        self.gmm = best_gmm 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def test_constants():
    assert chainerx.Inf is numpy.Inf
    assert chainerx.Infinity is numpy.Infinity
    assert chainerx.NAN is numpy.NAN
    assert chainerx.NINF is numpy.NINF
    assert chainerx.NZERO is numpy.NZERO
    assert chainerx.NaN is numpy.NaN
    assert chainerx.PINF is numpy.PINF
    assert chainerx.PZERO is numpy.PZERO
    assert chainerx.e is numpy.e
    assert chainerx.euler_gamma is numpy.euler_gamma
    assert chainerx.inf is numpy.inf
    assert chainerx.infty is numpy.infty
    assert chainerx.nan is numpy.nan
    assert chainerx.newaxis is numpy.newaxis
    assert chainerx.pi is numpy.pi 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def weiss_retinex(image, multi_images, mask, threshold, L1=False):
    multi_images = np.clip(multi_images, 3., np.infty)
    log_multi_images = np.log(multi_images)

    i_y_all, i_x_all = poisson.get_gradients(log_multi_images)
    r_y = np.median(i_y_all, axis=2)
    r_x = np.median(i_x_all, axis=2)

    r_y *= (np.abs(r_y) > threshold)
    r_x *= (np.abs(r_x) > threshold)
    if L1:
        log_refl = poisson.solve_L1(r_y, r_x, mask)
    else:
        log_refl = poisson.solve(r_y, r_x, mask)
    refl = np.where(mask, np.exp(log_refl), 0.)
    shading = np.where(mask, image / refl, 0.)

    return shading, refl





#################### Wrapper classes for experiments ########################### 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def assert_mpa_identical(mpa1, mpa2, decimal=np.infty):
    """Verify that two MPAs are complety identical
    """
    assert len(mpa1) == len(mpa2)
    assert mpa1.canonical_form == mpa2.canonical_form
    assert mpa1.dtype == mpa2.dtype

    for i, lten1, lten2 in zip(it.count(), mpa1.lt, mpa2.lt):
        if decimal is np.infty:
            assert_array_equal(lten1, lten2,
                               err_msg='mismatch in lten {}'.format(i))
        else:
            assert_array_almost_equal(lten1, lten2, decimal=decimal,
                                      err_msg='mismatch in lten {}'.format(i))
    # TODO: We should make a comprehensive comparison between `mpa1`
    # and `mpa2`.  Are we missing other things? 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def bsoid_hdbscan(umap_embeddings, hdbscan_params=HDBSCAN_PARAMS):
    """
    Trains HDBSCAN (unsupervised) given learned UMAP space
    :param umap_embeddings: 2D array, embedded UMAP space
    :param hdbscan_params: dict, HDBSCAN params in GLOBAL_CONFIG
    :return assignments: HDBSCAN assignments
    """
    highest_numulab = -np.infty
    numulab = []
    min_cluster_range = range(6, 21)
    logging.info('Running HDBSCAN on {} instances in {} D space...'.format(*umap_embeddings.shape))
    for min_c in min_cluster_range:
        trained_classifier = hdbscan.HDBSCAN(prediction_data=True,
                                             min_cluster_size=int(round(0.001 * min_c * umap_embeddings.shape[0])),
                                             **hdbscan_params).fit(umap_embeddings)
        numulab.append(len(np.unique(trained_classifier.labels_)))
        if numulab[-1] > highest_numulab:
            logging.info('Adjusting minimum cluster size to maximize cluster number...')
            highest_numulab = numulab[-1]
            best_clf = trained_classifier
    assignments = best_clf.labels_
    soft_clusters = hdbscan.all_points_membership_vectors(best_clf)
    soft_assignments = np.argmax(soft_clusters, axis=1)
    logging.info('Done predicting labels for {} instances in {} D space...'.format(*umap_embeddings.shape))
    return assignments, soft_clusters, soft_assignments 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def train_gmm(X, max_iter, tol, means, covariances):
    xp = cupy.get_array_module(X)
    lower_bound = -np.infty
    converged = False
    weights = xp.array([0.5, 0.5], dtype=np.float32)
    inv_cov = 1 / xp.sqrt(covariances)

    for n_iter in range(max_iter):
        prev_lower_bound = lower_bound
        log_prob_norm, log_resp = e_step(X, inv_cov, means, weights)
        weights, means, covariances = m_step(X, xp.exp(log_resp))
        inv_cov = 1 / xp.sqrt(covariances)
        lower_bound = log_prob_norm
        change = lower_bound - prev_lower_bound
        if abs(change) < tol:
            converged = True
            break

    if not converged:
        print('Failed to converge. Increase max-iter or tol.')

    return inv_cov, means, weights, covariances 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def test_sweepequalmatrix(self):
    for type in classes:
      self.swparams['collocation_class'] = type
      step, level, problem, nnodes = self.setupLevelStepProblem()
      step.levels[0].sweep.predict()
      u0full = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])

      # Perform node-to-node SDC sweep
      level.sweep.update_nodes()

      lambdas = [ problem.params.lambda_f[0] , problem.params.lambda_s[0] ]
      LHS, RHS = level.sweep.get_scalar_problems_sweeper_mats( lambdas = lambdas )

      unew = np.linalg.inv(LHS).dot( u0full + RHS.dot(u0full) )
      usweep = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])
      assert np.linalg.norm(unew - usweep, np.infty)<1e-14, "Single SDC sweeps in matrix and node-to-node formulation yield different results"

  #
  # Make sure the implemented update formula matches the matrix update formula
  # 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def test_updateformula(self):
    for type in classes:
      self.swparams['collocation_class'] = type
      step, level, problem, nnodes = self.setupLevelStepProblem()
      level.sweep.predict()
      u0full = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])

      # Perform update step in sweeper
      level.sweep.update_nodes()
      ustages = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])
      # Compute end value through provided function
      level.sweep.compute_end_point()
      uend_sweep = level.uend.values
      # Compute end value from matrix formulation
      if level.sweep.params.do_coll_update:
        uend_mat   = self.pparams['u0'] + step.dt*level.sweep.coll.weights.dot(ustages*(problem.params.lambda_s[0] + problem.params.lambda_f[0]))
      else:
        uend_mat = ustages[-1]
      assert np.linalg.norm(uend_sweep - uend_mat, np.infty)<1e-14, "Update formula in sweeper gives different result than matrix update formula"


  #
  # Compute the exact collocation solution by matrix inversion and make sure it is a fixed point
  # 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def test_updateformula_no_coll_update(self):
    for type in classes:
      self.swparams['collocation_class'] = type
      self.swparams['do_coll_update'] = False
      step, level, problem, nnodes = self.setupLevelStepProblem()
      # if type of nodes does not have right endpoint as quadrature nodes, cannot set do_coll_update to False and perform this test
      if not level.sweep.coll.right_is_node: break
      level.sweep.predict()
      u0full = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])

      # Perform update step in sweeper
      level.sweep.update_nodes()
      ustages = np.array([ level.u[l].values.flatten() for l in range(1,nnodes+1) ])

      # Compute end value through provided function
      level.sweep.compute_end_point()
      uend_sweep = level.uend.values
      # Compute end value from matrix formulation
      q = np.zeros(nnodes)
      q[nnodes-1] = 1.0
      uend_mat   = q.dot(ustages)
      assert np.linalg.norm(uend_sweep - uend_mat, np.infty)<1e-14, "For do_coll_update=False, update formula in sweeper gives different result than matrix update formula with q=(0,..,0,1)" 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def cal_Informativeness(self, label_index, unlabel_index, models):
        """

        Returns:
        Info : 2d array-like 
        shape [n_unlabel_samples, n_classes]
        Informativeness of each unlabel samples
        """
        Infor = np.zeros((self.n_samples, self.n_classes))
        Uncertainty = self.cal_uncertainty(label_index, unlabel_index, models)
        label_mat = label_index.get_matrix_mask((self.n_samples, self.n_classes), sparse=False)
        unlabel_mat = unlabel_index.get_matrix_mask((self.n_samples, self.n_classes), sparse=False)
        for j in np.arange(self.n_classes):
            j_unlabel = np.where(unlabel_mat[:, j] == 1)[0]
            j_label = np.where(unlabel_mat[:, j] != 1)[0]
            for i in j_unlabel:
                flag = self.cal_relevance(i, j, label_index, models, k=5)
                if flag == 1:
                    Infor[i][j] = Uncertainty[i][j] * 2
                elif flag == -1:
                    Infor[i][j] = Uncertainty[i][j] + self.cal_Udes(i, j, Uncertainty)
            Infor[j_label][j] = -np.infty
        return Infor 

# 需要导入模块: import numpy [as 别名]
# 或者: from numpy import infty [as 别名]
def cal_uncertainty(self, target, models):
        """Calculate the uncertainty.
        target: unlabel_martix
        """
        Uncertainty = np.zeros([self.n_samples, self.n_classes])
        # unlabel_data = self.X[unlabel_index, :]
        for j in np.arange(self.n_classes):
            model = models[j]
            j_target = target[:, j]
            j_label = np.where(j_target != 1)
            j_unlabel = np.where(j_target == 1)
            for i in j_unlabel[0]:
                d_v = model.decision_function([self.X[i]])
                Uncertainty[i][j] = np.abs(1 / d_v)
            Uncertainty[j_label, j] = -np.infty
        return Uncertainty