Python numpy.eye() Examples

def betaseries_file(tmpdir_factory,
                    deriv_betaseries_fname=deriv_betaseries_fname):
    bfile = tmpdir_factory.mktemp("beta").ensure(deriv_betaseries_fname)
    np.random.seed(3)
    num_trials = 40
    tgt_corr = 0.1
    bs1 = np.random.rand(num_trials)
    # create another betaseries with a target correlation
    bs2 = minimize(lambda x: abs(tgt_corr - pearsonr(bs1, x)[0]),
                   np.random.rand(num_trials)).x

    # two identical beta series
    bs_data = np.array([[[bs1, bs2]]])

    # the nifti image
    bs_img = nib.Nifti1Image(bs_data, np.eye(4))
    bs_img.to_filename(str(bfile))

    return bfile 

def time_align_visualize(alignments, time, y, namespace='time_align'):
    plt.figure()
    heat = np.flip(alignments + alignments.T +
                   np.eye(alignments.shape[0]), axis=0)
    sns.heatmap(heat, cmap="YlGnBu", vmin=0, vmax=1)
    plt.savefig(namespace + '_heatmap.svg')

    G = nx.from_numpy_matrix(alignments)
    G = nx.maximum_spanning_tree(G)

    pos = {}
    for i in range(len(G.nodes)):
        pos[i] = np.array([time[i], y[i]])

    mst_edges = set(nx.maximum_spanning_tree(G).edges())
    
    weights = [ G[u][v]['weight'] if (not (u, v) in mst_edges) else 8
                for u, v in G.edges() ]
    
    plt.figure()
    nx.draw(G, pos, edges=G.edges(), width=10)
    plt.ylim([-1, 1])
    plt.savefig(namespace + '.svg') 

def classical_mds(self, D):
        ''' 
        Classical multidimensional scaling

        Parameters
        ----------
        D : square 2D ndarray
            Euclidean Distance Matrix (matrix containing squared distances between points
        '''

        # Apply MDS algorithm for denoising
        n = D.shape[0]
        J = np.eye(n) - np.ones((n,n))/float(n)
        G = -0.5*np.dot(J, np.dot(D, J))

        s, U = np.linalg.eig(G)

        # we need to sort the eigenvalues in decreasing order
        s = np.real(s)
        o = np.argsort(s)
        s = s[o[::-1]]
        U = U[:,o[::-1]]

        S = np.diag(s)[0:self.dim,:]
        self.X = np.dot(np.sqrt(S),U.T) 

def output_shrink(K, L):
    """
    shrink the convolution output to half the size.
    used when both the annihilating filter and the uniform samples of sinusoids satisfy
    Hermitian symmetric.
    :param K: the annihilating filter size: K + 1
    :param L: length of the (complex-valued) b vector
    :return:
    """
    out_len = L - K
    if out_len % 2 == 0:
        half_out_len = np.int(out_len / 2.)
        mtx_r = np.hstack((np.eye(half_out_len),
                           np.zeros((half_out_len, half_out_len))))
        mtx_i = mtx_r
    else:
        half_out_len = np.int((out_len + 1) / 2.)
        mtx_r = np.hstack((np.eye(half_out_len),
                           np.zeros((half_out_len, half_out_len - 1))))
        mtx_i = np.hstack((np.eye(half_out_len - 1),
                           np.zeros((half_out_len - 1, half_out_len))))
    return linalg.block_diag(mtx_r, mtx_i) 

def mtx_updated_G(phi_recon, M, mtx_amp2visi_ri, mtx_fri2visi_ri):
    """
    Update the linear transformation matrix that links the FRI sequence to the
    visibilities by using the reconstructed Dirac locations.
    :param phi_recon: the reconstructed Dirac locations (azimuths)
    :param M: the Fourier series expansion is between -M to M
    :param p_mic_x: a vector that contains microphones' x-coordinates
    :param p_mic_y: a vector that contains microphones' y-coordinates
    :param mtx_freq2visi: the linear mapping from Fourier series to visibilities
    :return:
    """
    L = 2 * M + 1
    ms_half = np.reshape(np.arange(-M, 1, step=1), (-1, 1), order='F')
    phi_recon = np.reshape(phi_recon, (1, -1), order='F')
    mtx_amp2freq = np.exp(-1j * ms_half * phi_recon)  # size: (M + 1) x K
    mtx_amp2freq_ri = np.vstack((mtx_amp2freq.real, mtx_amp2freq.imag[:-1, :]))  # size: (2M + 1) x K
    mtx_fri2amp_ri = linalg.lstsq(mtx_amp2freq_ri, np.eye(L))[0]
    # projection mtx_freq2visi to the null space of mtx_fri2amp
    mtx_null_proj = np.eye(L) - np.dot(mtx_fri2amp_ri.T,
                                       linalg.lstsq(mtx_fri2amp_ri.T, np.eye(L))[0])
    G_updated = np.dot(mtx_amp2visi_ri, mtx_fri2amp_ri) + \
                np.dot(mtx_fri2visi_ri, mtx_null_proj)
    return G_updated 

def _N(self,s,r):
        """
        Lagrange's interpolate function
        params:
            s,r:natural position of evalue point.2-array.
        returns:
            2x(2x4) shape function matrix.
        """
        la1=(1-s)/2
        la2=(1+s)/2
        lb1=(1-r)/2
        lb2=(1+r)/2
        N1=la1*lb1
        N2=la1*lb2
        N3=la2*lb1
        N4=la2*lb2

        N=np.hstack(N1*np.eye(2),N2*np.eye(2),N3*np.eye(2),N4*np.eye(2))
        return N 

def __init__(self):
        """Initialize variable used by Kalman Filter class
        Args:
            None
        Return:
            None
        """
        self.dt = 0.005  # delta time

        self.A = np.array([[1, 0], [0, 1]])  # matrix in observation equations
        self.u = np.zeros((2, 1))  # previous state vector

        # (x,y) tracking object center
        self.b = np.array([[0], [255]])  # vector of observations

        self.P = np.diag((3.0, 3.0))  # covariance matrix
        self.F = np.array([[1.0, self.dt], [0.0, 1.0]])  # state transition mat

        self.Q = np.eye(self.u.shape[0])  # process noise matrix
        self.R = np.eye(self.b.shape[0])  # observation noise matrix
        self.lastResult = np.array([[0], [255]]) 

def test_output():
    shape = (2,2)
    ones = mx.nd.ones(shape)
    zeros = mx.nd.zeros(shape)
    out = mx.nd.zeros(shape)
    mx.nd.ones(shape, out=out)
    assert_almost_equal(out.asnumpy(), ones.asnumpy())
    mx.nd.zeros(shape, out=out)
    assert_almost_equal(out.asnumpy(), zeros.asnumpy())
    mx.nd.full(shape, 2, out=out)
    assert_almost_equal(out.asnumpy(), ones.asnumpy() * 2)
    arange_out = mx.nd.arange(0, 20, dtype='int64')
    assert_almost_equal(arange_out.asnumpy(), np.arange(0, 20))
    N_array = np.random.randint(1, high=8, size=10)
    M_array = np.random.randint(1, high=8, size=10)
    k_array = np.random.randint(-10, high=10, size=10)
    for i in range(10):
        N = N_array[i]
        M = M_array[i]
        k = k_array[i]
        assert_almost_equal(np.eye(N, M, k), mx.nd.eye(N, M, k).asnumpy())
        assert_almost_equal(np.eye(N, k=k), mx.nd.eye(N, k=k).asnumpy()) 

def set_camera(self, fov_vertical, z_near, z_far, aspect):
    width = 2*np.tan(np.deg2rad(fov_vertical)/2.0)*z_near*aspect;
    height = 2*np.tan(np.deg2rad(fov_vertical)/2.0)*z_near;
    egl_program = self.egl_program
    c = np.eye(4, dtype=np.float32)
    c[3,3] = 0
    c[3,2] = -1
    c[2,2] = -(z_near+z_far)/(z_far-z_near)
    c[2,3] = -2.0*(z_near*z_far)/(z_far-z_near)
    c[0,0] = 2.0*z_near/width
    c[1,1] = 2.0*z_near/height
    c = c.T
    
    projection_matrix_o = glGetUniformLocation(egl_program, 'uProjectionMatrix')
    projection_matrix = np.eye(4, dtype=np.float32)
    projection_matrix[...] = c
    projection_matrix = np.reshape(projection_matrix, (-1))
    glUniformMatrix4fv(projection_matrix_o, 1, GL_FALSE, projection_matrix) 

def rotate_camera_to_point_at(up_from, lookat_from, up_to, lookat_to):
  inputs = [up_from, lookat_from, up_to, lookat_to]
  for i in range(4):
    inputs[i] = normalize(np.array(inputs[i]).reshape((-1,)))
  up_from, lookat_from, up_to, lookat_to = inputs
  r1 = r_between(lookat_from, lookat_to)

  new_x = np.dot(r1, np.array([1, 0, 0]).reshape((-1, 1))).reshape((-1))
  to_x = normalize(np.cross(lookat_to, up_to))
  angle = np.arccos(np.dot(new_x, to_x))
  if angle > ANGLE_EPS:
    if angle < np.pi - ANGLE_EPS:
      ax = normalize(np.cross(new_x, to_x))
      flip = np.dot(lookat_to, ax)
      if flip > 0:
        r2 = get_r_matrix(lookat_to, angle)
      elif flip < 0:
        r2 = get_r_matrix(lookat_to, -1. * angle)
    else:
      # Angle of rotation is too close to 180 degrees, direction of rotation
      # does not matter.
      r2 = get_r_matrix(lookat_to, angle)
  else:
    r2 = np.eye(3)
  return np.dot(r2, r1) 

def test_givens_inverse():
    r"""
    The Givens rotation in OpenFermion is defined as

    .. math::

        \begin{pmatrix}
            \cos(\theta) & -e^{i \varphi} \sin(\theta) \\
            \sin(\theta) &     e^{i \varphi} \cos(\theta)
        \end{pmatrix}.

    confirm numerically its hermitian conjugate is it's inverse
    """
    a = numpy.random.random() + 1j * numpy.random.random()
    b = numpy.random.random() + 1j * numpy.random.random()
    ab_rotation = givens_matrix_elements(a, b, which='right')

    assert numpy.allclose(ab_rotation.dot(numpy.conj(ab_rotation).T),
                          numpy.eye(2))
    assert numpy.allclose(numpy.conj(ab_rotation).T.dot(ab_rotation),
                          numpy.eye(2)) 

def test_circuit_generation_and_accuracy():
    for dim in range(2, 10):
        qubits = cirq.LineQubit.range(dim)
        u_generator = numpy.random.random(
            (dim, dim)) + 1j * numpy.random.random((dim, dim))
        u_generator = u_generator - numpy.conj(u_generator).T
        assert numpy.allclose(-1 * u_generator, numpy.conj(u_generator).T)

        unitary = scipy.linalg.expm(u_generator)
        circuit = cirq.Circuit()
        circuit.append(optimal_givens_decomposition(qubits, unitary))

        fermion_generator = QubitOperator(()) * 0.0
        for i, j in product(range(dim), repeat=2):
            fermion_generator += jordan_wigner(
                FermionOperator(((i, 1), (j, 0)), u_generator[i, j]))

        true_unitary = scipy.linalg.expm(
            get_sparse_operator(fermion_generator).toarray())
        assert numpy.allclose(true_unitary.conj().T.dot(true_unitary),
                              numpy.eye(2 ** dim, dtype=complex))

        test_unitary = cirq.unitary(circuit)
        assert numpy.isclose(
            abs(numpy.trace(true_unitary.conj().T.dot(test_unitary))), 2 ** dim) 

def energy_from_opdm(opdm, constant, one_body_tensor, two_body_tensor):
    """
    Evaluate the energy of an opdm assuming the 2-RDM is opdm ^ opdm

    :param opdm: single spin-component of the full spin-orbital opdm.
    :param constant: constant shift to the Hamiltonian. Commonly this is the
                     nuclear repulsion energy.
    :param one_body_tensor: spatial one-body integrals
    :param two_body_tensor: spatial two-body integrals
    :return:
    """
    spin_opdm = np.kron(opdm, np.eye(2))
    spin_tpdm = 2 * wedge(spin_opdm, spin_opdm, (1, 1), (1, 1))
    molecular_hamiltonian = generate_hamiltonian(constant=constant,
                                                 one_body_integrals=one_body_tensor,
                                                 two_body_integrals=two_body_tensor)
    rdms = InteractionRDM(spin_opdm, spin_tpdm)
    return rdms.expectation(molecular_hamiltonian).real 

def netflix(es, ps, e0, l=.0001):
    """Combine predictions with the optimal weights to minimize RMSE.

    Args:
        es (list of float): RMSEs of predictions
        ps (list of np.array): predictions
        e0 (float): RMSE of all zero prediction
        l (float): lambda as in the ridge regression

    Returns:
        (tuple):

            - (np.array): ensemble predictions
            - (np.array): weights for input predictions
    """
    m = len(es)
    n = len(ps[0])

    X = np.stack(ps).T
    pTy = .5 * (n * e0**2 + (X**2).sum(axis=0) - n * np.array(es)**2)

    w = np.linalg.pinv(X.T.dot(X) + l * n * np.eye(m)).dot(pTy)

    return X.dot(w), w 

def fit(self, Xs, Xt):
        '''
        Fit source and target using KMM (compute the coefficients)
        :param Xs: ns * dim
        :param Xt: nt * dim
        :return: Coefficients (Pt / Ps) value vector (Beta in the paper)
        '''
        ns = Xs.shape[0]
        nt = Xt.shape[0]
        if self.eps == None:
            self.eps = self.B / np.sqrt(ns)
        K = kernel(self.kernel_type, Xs, None, self.gamma)
        kappa = np.sum(kernel(self.kernel_type, Xs, Xt, self.gamma) * float(ns) / float(nt), axis=1)

        K = matrix(K)
        kappa = matrix(kappa)
        G = matrix(np.r_[np.ones((1, ns)), -np.ones((1, ns)), np.eye(ns), -np.eye(ns)])
        h = matrix(np.r_[ns * (1 + self.eps), ns * (self.eps - 1), self.B * np.ones((ns,)), np.zeros((ns,))])

        sol = solvers.qp(K, -kappa, G, h)
        beta = np.array(sol['x'])
        return beta 

def fit(self, Xs, Xt):
        '''
        Transform Xs and Xt
        :param Xs: ns * n_feature, source feature
        :param Xt: nt * n_feature, target feature
        :return: Xs_new and Xt_new after TCA
        '''
        X = np.hstack((Xs.T, Xt.T))
        X /= np.linalg.norm(X, axis=0)
        m, n = X.shape
        ns, nt = len(Xs), len(Xt)
        e = np.vstack((1 / ns * np.ones((ns, 1)), -1 / nt * np.ones((nt, 1))))
        M = e * e.T
        M = M / np.linalg.norm(M, 'fro')
        H = np.eye(n) - 1 / n * np.ones((n, n))
        K = kernel(self.kernel_type, X, None, gamma=self.gamma)
        n_eye = m if self.kernel_type == 'primal' else n
        a, b = np.linalg.multi_dot([K, M, K.T]) + self.lamb * np.eye(n_eye), np.linalg.multi_dot([K, H, K.T])
        w, V = scipy.linalg.eig(a, b)
        ind = np.argsort(w)
        A = V[:, ind[:self.dim]]
        Z = np.dot(A.T, K)
        Z /= np.linalg.norm(Z, axis=0)
        Xs_new, Xt_new = Z[:, :ns].T, Z[:, ns:].T
        return Xs_new, Xt_new 

def quickSave(img, wmh, gif, n):
    nft_img = nib.Nifti1Image(img.squeeze(), np.eye(4))
    nib.save(nft_img, n + '_img.nii.gz')
    nft_img = nib.Nifti1Image(wmh.squeeze(), np.eye(4))
    nib.save(nft_img, n + '_wmh.nii.gz')
    if gif is not None:
        nft_img = nib.Nifti1Image(gif.squeeze(), np.eye(4))
        nib.save(nft_img, n + '_gif.nii.gz')

#------------------------------------------------------------------------------
#END OF SLICES PREPROCESSING
#------------------------------------------------------------------------------

#------------------------------------------------------------------------------
#3D PREPROCESSING
#------------------------------------------------------------------------------
#go through every 3D object from training set and every patch of size NxNxN
#save resulting 3D volumes in one of the two folders based on what the center pixel of the image is 

def invertFast(A, d):
	"""
	given an array A of shape d x k and a d x 1 vector d, computes (A * A.T + diag(d)) ^{-1}
	Checked.
	"""
	assert(A.shape[0] == d.shape[0])
	assert(d.shape[1] == 1)

	k = A.shape[1]
	A = np.array(A)
	d_vec = np.array(d)
	d_inv = np.array(1 / d_vec[:, 0])

	inv_d_squared = np.dot(np.atleast_2d(d_inv).T, np.atleast_2d(d_inv))
	M = np.diag(d_inv) - inv_d_squared * np.dot(np.dot(A, np.linalg.inv(np.eye(k, k) + np.dot(A.T, mult_diag(d_inv, A)))), A.T)

	return M 

def get_training_data(num_samples):
    """Generates some training data."""

    # As (x, y) Cartesian coordinates.
    x = np.random.randint(0, 2, size=(num_samples, 2))

    y = x[:, 0] + 2 * x[:, 1]  # 2-digit binary to integer.
    y = np.cast['int32'](y)

    x = np.cast['float32'](x) * 1.6 - 0.8  # Scales to [-1, 1].
    x += np.random.uniform(-0.1, 0.1, size=x.shape)

    y_ohe = np.cast['float32'](np.eye(4)[y])
    y = np.cast['float32'](np.expand_dims(y, -1))

    return x, y, y_ohe 

def get_mnist_data(binarize=False):
    """Puts the MNIST data in the right format."""

    (X_train, y_train), (X_test, y_test) = mnist.load_data()

    if binarize:
        X_test = np.where(X_test >= 10, 1, -1)
        X_train = np.where(X_train >= 10, 1, -1)
    else:
        X_train = (X_train.astype(np.float32) - 127.5) / 127.5
        X_test = (X_test.astype(np.float32) - 127.5) / 127.5

    X_train = np.expand_dims(X_train, axis=-1)
    X_test = np.expand_dims(X_test, axis=-1)

    y_train = np.eye(10)[y_train]
    y_test = np.eye(10)[y_test]

    return (X_train, y_train), (X_test, y_test) 

def reg_cov(self, X):
        """
        Regularize covariance matrix until non-singular.

        Parameters
        ----------
        C : array
            square symmetric covariance matrix.

        Returns
        -------
        C : array
            regularized covariance matrix.

        """
        # Compute mean of data
        muX = np.mean(X, axis=0, keepdims=1)

        # Compute covariance matrix without regularization
        SX = np.cov((X - muX).T)

        # Initialize regularization parameter
        reg = 1e-6

        # Keep going until non-singular
        while not self.is_pos_def(SX):

            # Compute covariance matrix with regularization
            SX = np.cov((X - muX).T) + reg*np.eye(X.shape[1])

            # Increment reg
            reg *= 10

        # Report regularization
        print('Final regularization parameter = {}'.format(reg))

        return SX 

def reg_cov(self, X):
        """
        Regularize covariance matrix until non-singular.

        Parameters
        ----------
        C : array
            square symmetric covariance matrix.

        Returns
        -------
        C : array
            regularized covariance matrix.

        """
        # Number of data points
        N = X.shape[0]

        # Compute mean of data
        muX = np.mean(X, axis=0, keepdims=1)

        # Compute covariance matrix without regularization
        SX = np.dot((X - muX).T, (X - muX)) / N

        # Initialize regularization parameter
        reg = 1e-6

        # Keep going until non-singular
        while not self.is_pos_def(SX):

            # Compute covariance matrix with regularization
            SX = np.dot((X - muX).T, (X - muX)) / N + reg*np.eye(X.shape[1])

            # Increment reg
            reg *= 10

        # Report regularization
        print('Final regularization parameter = {}'.format(reg))

        return SX 

def eye(size, dtype=None, name=None):
    '''Instantiates an identity matrix.
    '''
    if dtype is None:
        dtype = floatx()
    return variable(np.eye(size), dtype, name) 

def extract_off_diag(mtx):
    """
    extract off diagonal entries in mtx.
    The output vector is order in a column major manner.
    :param mtx: input matrix to extract the off diagonal entries
    :return:
    """
    # we transpose the matrix because the function np.extract will first flatten the matrix
    # withe ordering convention 'C' instead of 'F'!!
    extract_cond = np.reshape((1 - np.eye(*mtx.shape)).T.astype(bool), (-1, 1), order='F')
    return np.reshape(np.extract(extract_cond, mtx.T), (-1, 1), order='F') 

def hermitian_expan(half_vec_len):
    """
    expand a real-valued vector to a Hermitian symmetric vector.
    The input vector is a concatenation of the real parts with NON-POSITIVE indices and
    the imaginary parts with STRICTLY-NEGATIVE indices.
    :param half_vec_len: length of the first half vector
    :return:
    """
    D0 = np.eye(half_vec_len)
    D1 = np.vstack((D0, D0[1:, ::-1]))
    D2 = np.vstack((D0, -D0[1:, ::-1]))
    D2 = D2[:, :-1]
    return D1, D2 

def coef_expan_mtx(K):
    """
    expansion matrix for an annihilating filter of size K + 1
    :param K: number of Dirac. The filter size is K + 1
    :return:
    """
    if K % 2 == 0:
        D0 = np.eye(np.int(K / 2. + 1))
        D1 = np.vstack((D0, D0[1:, ::-1]))
        D2 = np.vstack((D0, -D0[1:, ::-1]))[:, :-1]
    else:
        D0 = np.eye(np.int((K + 1) / 2.))
        D1 = np.vstack((D0, D0[:, ::-1]))
        D2 = np.vstack((D0, -D0[:, ::-1]))
    return D1, D2 

def __init__(self,node_i, node_j, E, A, rho, name=None, mass='conc', tol=1e-6):
        """
        params:
            node_i,node_j: ends of link.
            E: elastic modulus
            A: section area
            rho: mass density
            mass: 'coor' as coordinate matrix or 'conc' for concentrated matrix
            tol: tolerance
        """
        super(Link,self).__init__(node_i,node_j,A,rho,6,name,mass)
        l=self._length
        K_data=(
            (E*A/l,(0,0)),
            (-E*A/l,(0,1)),
            (-E*A/l,(1,0)),
            (E*A/l,(1,1)),
        )
        m_data=(
            (1,(0,0)),
            (1,(1,1))*rho*A*l/2
        )
        data=[k[0] for k in K_data]
        row=[k[1][0] for k in K_data]
        col=[k[1][1] for k in K_data]
        self._Ke = spr.csr_matrix((data,(row,col)),shape=(12, 12))
        self._Me=spr.eye(2)*rho*A*l/2
        #force vector
        self._re =np.zeros((2,1)) 

def _N(self,s):
        """
        params:
            Hermite's interpolate function
            s:natural position of evalue point.
        returns:
            3x(3x4) shape function matrix.
        """
        N1=1-3*s**2+2*s**3
        N2=  3*s**2-2*s**3
        N3=s-2*s**2+s**3
        N4=   -s**2+s**3
        N=np.hstack([np.eye(3)*N1,np.eye(3)*N2,np.eye(3)*N3,np.eye(3)*N4])
        return N 

def __init__(self,node_i, node_j, node_k, node_l, t, E, mu, rho, name=None):
        """
        node_i,node_j,node_k: corners of triangle.
        t: thickness
        E: elastic modulus
        mu: Poisson ratio
        rho: mass density
        """
        super(Membrane4,self).__init__(node_i,node_j,node_k,node_l,t,E,mu,rho,8)
        self._t=t
        self._E=E
        self._mu=mu
        self._rho=rho

        elm_node_count=4
        node_dof=2
        
        Ke = quadpy.quadrilateral.integrate(
            lambda s: self._BtDB(s[0],s[1])*np.linalg.det(self._J(s[0],s[1])),
            quadpy.quadrilateral.rectangle_points([-1.0, 1.0], [-1.0, 1.0]),
            quadpy.quadrilateral.Stroud('C2 7-2')
            )
        Ke=sp.sparse.csr_matrix(Ke)
        row=[]
        col=[]
        for i in range(elm_node_count):
            row+=[a for a in range(i*node_dof,i*node_dof+node_dof)]
            col+=[a for a in range(i*6,i*6+node_dof)]
        data=[1]*(elm_node_count*node_dof)
        G=sp.sparse.csr_matrix((data,(row,col)),shape=(elm_node_count*node_dof,elm_node_count*6))
        self._Ke=G.transpose()*Ke*G
        #Concentrated mass matrix, may be wrong
        self._Me=G.transpose()*np.eye(node_dof*elm_node_count)*G*rho*self._area*t/4
        
        self._re =np.zeros((elm_node_count*6,1)) 

def attack(self, X, Y):
        """
        Perform the L_2 attack on the given images for the given targets.

        :param X: samples to generate advs
        :param Y: the original class labels
        If self.targeted is true, then the targets represents the target labels.
        If self.targeted is false, then targets are the original class labels.
        """
        nb_classes = Y.shape[1]

        # random select target class for targeted attack
        y_target = np.copy(Y)
        if self.TARGETED:
            for i in range(Y.shape[0]):
                current = int(np.argmax(Y[i]))
                target = np.random.choice(other_classes(nb_classes, current))
                y_target[i] = np.eye(nb_classes)[target]

        X_adv = np.zeros_like(X)
        for i in tqdm(range(0, X.shape[0], self.batch_size)):
            start = i
            end = i + self.batch_size
            end = np.minimum(end, X.shape[0])
            X_adv[start:end] = self.attack_batch(X[start:end], y_target[start:end])

        return X_adv