Python numpy.triu_indices方法代碼示例

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def _do_continue(self, algorithm):
        do_continue = self.default.do_continue(algorithm)

        # if the default says do not continue just follow that
        if not do_continue:
            return False

        # additionally check for degenerated simplex
        else:
            X = algorithm.pop.get("X")

            # degenerated simplex - get all edges and minimum and maximum length
            D = vectorized_cdist(X, X)
            val = D[np.triu_indices(len(X), 1)]
            min_e, max_e = val.min(), val.max()

            # either if the maximum length is very small or the ratio is degenerated
            is_degenerated = max_e < 1e-16 or min_e / max_e < 1e-16

            return not is_degenerated 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def __init__(self, features, using_cache=False, identity_init=True, eps=1e-3):
        super().__init__(features, using_cache)

        self.eps = eps

        self.lower_indices = np.tril_indices(features, k=-1)
        self.upper_indices = np.triu_indices(features, k=1)
        self.diag_indices = np.diag_indices(features)

        n_triangular_entries = ((features - 1) * features) // 2

        self.lower_entries = nn.Parameter(torch.zeros(n_triangular_entries))
        self.upper_entries = nn.Parameter(torch.zeros(n_triangular_entries))
        self.unconstrained_upper_diag = nn.Parameter(torch.zeros(features))

        self._initialize(identity_init) 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def ut_dense(preds_ut, diagonal_offset):
  """Construct dense prediction matrix from upper triangular."""
  ut_len, num_targets = preds_ut.shape

  # infer original sequence length
  seq_len = int(np.sqrt(2*ut_len + 0.25) - 0.5)
  seq_len += diagonal_offset

  # get triu indexes
  ut_indexes = np.triu_indices(seq_len, diagonal_offset)
  assert(len(ut_indexes[0]) == ut_len)

  # assign to dense matrix
  preds_dense = np.zeros(shape=(seq_len,seq_len,num_targets), dtype=preds_ut.dtype)
  preds_dense[ut_indexes] = preds_ut

  # symmetrize
  preds_dense += np.transpose(preds_dense, axes=[1,0,2])

  return preds_dense 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def pairwise_distance_np(feature, squared=False):
    """Computes the pairwise distance matrix in numpy.
    Args:
        feature: 2-D numpy array of size [number of data, feature dimension]
        squared: Boolean. If true, output is the pairwise squared euclidean
                 distance matrix; else, output is the pairwise euclidean distance matrix.
    Returns:
        pairwise_distances: 2-D numpy array of size
                            [number of data, number of data].
    """
    triu = np.triu_indices(feature.shape[0], 1)
    upper_tri_pdists = np.linalg.norm(feature[triu[1]] - feature[triu[0]], axis=1)
    if squared:
        upper_tri_pdists **= 2.
    num_data = feature.shape[0]
    pairwise_distances = np.zeros((num_data, num_data))
    pairwise_distances[np.triu_indices(num_data, 1)] = upper_tri_pdists
    # Make symmetrical.
    pairwise_distances = pairwise_distances + pairwise_distances.T - np.diag(
        pairwise_distances.diagonal())
    return pairwise_distances 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def _get_lvec(labels):
    """
    Constructs a label vector for an arbitrary number of labels
    Assumes that our model is quadratic in the labels

    Parameters
    ----------
    labels: numpy ndarray
        pivoted label values for one star

    Returns
    -------
    lvec: numpy ndarray
        label vector
    """
    nlabels = len(labels)
    # specialized to second-order model
    linear_terms = labels 
    quadratic_terms = np.outer(linear_terms, 
                               linear_terms)[np.triu_indices(nlabels)]
    lvec = np.hstack((linear_terms, quadratic_terms))
    return lvec 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def compare_ordered(vals, alpha):
    '''simple ordered sequential comparison of means

    vals : array_like
        means or rankmeans for independent groups

    incomplete, no return, not used yet
    '''
    vals = np.asarray(vals)
    alphaf = alpha  # Notation ?
    sortind = np.argsort(vals)
    pvals = vals[sortind]
    sortrevind = sortind.argsort()
    ntests = len(vals)
    #alphacSidak = 1 - np.power((1. - alphaf), 1./ntests)
    #alphacBonf = alphaf / float(ntests)
    v1, v2 = np.triu_indices(ntests, 1)
    #v1,v2 have wrong sequence
    for i in range(4):
        for j in range(4,i, -1):
            print(i,j) 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def _hessian_geom(self, params):
        exog = self.exog
        y = self.endog[:,None]
        mu = self.predict(params)[:,None]

        # for dl/dparams dparams
        dim = exog.shape[1]
        hess_arr = np.empty((dim, dim))
        const_arr = mu*(1+y)/(mu+1)**2
        for i in range(dim):
            for j in range(dim):
                if j > i:
                    continue
                hess_arr[i,j] = np.sum(-exog[:,i,None] * exog[:,j,None] *
                                       const_arr, axis=0)
        tri_idx = np.triu_indices(dim, k=1)
        hess_arr[tri_idx] = hess_arr.T[tri_idx]
        return hess_arr 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def _helmert_contrast(self, levels):
        n = len(levels)
        #http://www.ats.ucla.edu/stat/sas/webbooks/reg/chapter5/sasreg5.htm#HELMERT
        #contr = np.eye(n - 1)
        #int_range = np.arange(n - 1., 1, -1)
        #denom = np.repeat(int_range, np.arange(n - 2, 0, -1))
        #contr[np.tril_indices(n - 1, -1)] = -1. / denom

        #http://www.ats.ucla.edu/stat/r/library/contrast_coding.htm#HELMERT
        #contr = np.zeros((n - 1., n - 1))
        #int_range = np.arange(n, 1, -1)
        #denom = np.repeat(int_range[:-1], np.arange(n - 2, 0, -1))
        #contr[np.diag_indices(n - 1)] = (int_range - 1.) / int_range
        #contr[np.tril_indices(n - 1, -1)] = -1. / denom
        #contr = np.vstack((contr, -1./int_range))

        #r-like
        contr = np.zeros((n, n - 1))
        contr[1:][np.diag_indices(n - 1)] = np.arange(1, n)
        contr[np.triu_indices(n - 1)] = -1
        return contr 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def _diff_contrast(self, levels):
        nlevels = len(levels)
        contr = np.zeros((nlevels, nlevels-1))
        int_range = np.arange(1, nlevels)
        upper_int = np.repeat(int_range, int_range)
        row_i, col_i = np.triu_indices(nlevels-1)
        # we want to iterate down the columns not across the rows
        # it would be nice if the index functions had a row/col order arg
        col_order = np.argsort(col_i)
        contr[row_i[col_order],
              col_i[col_order]] = (upper_int-nlevels)/float(nlevels)
        lower_int = np.repeat(int_range, int_range[::-1])
        row_i, col_i = np.tril_indices(nlevels-1)
        # we want to iterate down the columns not across the rows
        col_order = np.argsort(col_i)
        contr[row_i[col_order]+1, col_i[col_order]] = lower_int/float(nlevels)
        return contr 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def pairwise_distances(feature, squared=True):
  """Computes the pairwise distance matrix in numpy.

  Args:
    feature: 2-D numpy array of size [number of data, feature dimension]
    squared: Boolean. If true, output is the pairwise squared euclidean
      distance matrix; else, output is the pairwise euclidean distance matrix.

  Returns:
    pdists: 2-D numpy array of size
      [number of data, number of data].
  """
  triu = np.triu_indices(feature.shape[0], 1)
  upper_tri_pdists = np.linalg.norm(feature[triu[1]] - feature[triu[0]], axis=1)
  if squared:
    upper_tri_pdists **= 2.
  num_data = feature.shape[0]
  pdists = np.zeros((num_data, num_data))
  pdists[np.triu_indices(num_data, 1)] = upper_tri_pdists
  # Make symmetrical.
  pdists = pdists + pdists.T - np.diag(
      pdists.diagonal())
  return pdists 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def test_idxiter():
    n_channels = data.shape[0]
    # Upper-triangular part, including diag
    idx0, idx1 = np.triu_indices(n_channels)
    triu_indices = np.array([np.arange(idx0.size), idx0, idx1])
    triu_indices2 = np.array(list(_idxiter(n_channels, include_diag=True)))
    # Upper-triangular part, without diag
    idx2, idx3 = np.triu_indices(n_channels, 1)
    triu_indices_nodiag = np.array([np.arange(idx2.size), idx2, idx3])
    triu_indices2_nodiag = np.array(list(_idxiter(n_channels,
                                                  include_diag=False)))
    assert_almost_equal(triu_indices, triu_indices2.transpose())
    assert_almost_equal(triu_indices_nodiag, triu_indices2_nodiag.transpose())
    # Upper and lower-triangular parts, without diag
    expected = [(i, j) for _, (i, j) in
                enumerate(np.ndindex((n_channels, n_channels))) if i != j]
    assert_equal(np.array([(i, j) for _, i, j in _idxiter(n_channels,
                                                          triu=False)]),
                 expected) 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def pairwise_distance_np(feature, squared=False):
  """Computes the pairwise distance matrix in numpy.

  Args:
    feature: 2-D numpy array of size [number of data, feature dimension]
    squared: Boolean. If true, output is the pairwise squared euclidean
      distance matrix; else, output is the pairwise euclidean distance matrix.

  Returns:
    pairwise_distances: 2-D numpy array of size
      [number of data, number of data].
  """
  triu = np.triu_indices(feature.shape[0], 1)
  upper_tri_pdists = np.linalg.norm(feature[triu[1]] - feature[triu[0]], axis=1)
  if squared:
    upper_tri_pdists **= 2.
  num_data = feature.shape[0]
  pairwise_distances = np.zeros((num_data, num_data))
  pairwise_distances[np.triu_indices(num_data, 1)] = upper_tri_pdists
  # Make symmetrical.
  pairwise_distances = pairwise_distances + pairwise_distances.T - np.diag(
      pairwise_distances.diagonal())
  return pairwise_distances 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def from_tri_2_sym(tri, dim):
    """convert a upper triangular matrix in 1D format
       to 2D symmetric matrix


    Parameters
    ----------

    tri: 1D array
        Contains elements of upper triangular matrix

    dim : int
        The dimension of target matrix.


    Returns
    -------

    symm : 2D array
        Symmetric matrix in shape=[dim, dim]
    """
    symm = np.zeros((dim, dim))
    symm[np.triu_indices(dim)] = tri
    return symm 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def compute_rand_index(emb, labels):
    """
    https://en.wikipedia.org/wiki/Rand_index
    """
    n = len(emb)
    k = np.unique(labels).size

    m = KMeans(k)
    m.fit(emb)
    emb_labels = m.predict(emb)

    agreements = 0
    for i, j in zip(*np.triu_indices(n, 1)):
        emb_same = emb_labels[i] == emb_labels[j]
        gt_same = labels[i] == labels[j]

        if emb_same == gt_same:
            agreements += 1

    return float(agreements) / (n * (n-1) / 2) 

# 需要導入模塊: import numpy [as 別名]
# 或者: from numpy import triu_indices [as 別名]
def predict(self, m, s=None, t=None):
        D = m.shape[0]
        if t is not None:
            self.t = t
        t = self.t

        u, z, I, L = self.u_nominal, self.z_nominal, self.I_, self.L_

        if s is None:
            s = np.zeros((D, D))

        # construct flattened state covariance vector
        z_t = tt_.concatenate([m.flatten(), s[self.triu_indices]])
        # compute control
        u_t = u[t] + I[t] + L[t].dot(z_t - z[t])

        # limit the controller output
        #u_t = tt.clip(u_t, -self.maxU,  self.maxU)

        U = u_t.shape[0]
        self.t += 1
        return u_t, tt_.zeros((U, U)), tt_.zeros((D, U))