Permalink
Commits on Dec 12, 2010
  1. REL: set version to 0.9.0b1

    rgommers committed Dec 12, 2010
  2. BUG: 3K: fix integer division and list.sort issues in kendalltau.

    rgommers committed Dec 12, 2010
    Backported from r7006.
Commits on Dec 8, 2010
  1. TST: change criterion for fmin_ncg test. Closes #1323.

    rgommers committed Dec 8, 2010
    The number of gradient evaluations has a slight platform dependence due to
    floating-point comparison differences. Number of calls is 18 on most platforms,
    but has been reported as 16 in the ticket and on-list, with Python 2.7 and 2.5
    respectively.
Commits on Dec 5, 2010
Commits on Dec 4, 2010
  1. ENH: sparse/arpack: rewrite the sparse svd routine to handle complex …

    pv committed Dec 4, 2010
    …matrices
    
    The ARPACK user guide implies that the only way to compute eigenvalues
    of Hermitian matrices is to use zneupd, as there is no special support
    for hermitian matrices.
  2. ENH: sparse/arpack: Rename routines eigen* -> eigs* to avoid conflict…

    pv committed Dec 4, 2010
    … with module name
  3. BUG: sparse/arpack: fix bug in return value extraction from dneupd, w…

    pv committed Dec 4, 2010
    …hen got less eigenvalues than desired
  4. BUG: sparse/arpack: fix ARPACK error handling

    pv committed Dec 4, 2010
    Make exceptions instances of ArpackError. Raise ArpackNoConvergence
    always when no convergence is obtained.
  5. ENH: sparse: un-deprecate spmatrix.dot --- ndarrays also have the .do…

    pv committed Dec 4, 2010
    …t() method meaning the matrix product, so sparse matrices should also have it
Commits on Dec 3, 2010
  1. BUG: stats: Eliminate the 'TINY' hack in pearsonr, so perfect correla…

    warren.weckesser committed Dec 3, 2010
    …tions return probability=0. Also allow for numerical errors that could make the computed value of r be less than -1.
Commits on Dec 1, 2010
  1. BUG: ndimage: correctly deal with 'labels' data types that are compat…

    warren.weckesser committed Dec 1, 2010
    …ible with bincount (tickets #1254 and #1242).  Use numpy.unique instead of numpy.unique1d.
Commits on Nov 29, 2010
  1. TST: signal: make the decimal precision of the complex tests of the c…

    warren.weckesser committed Nov 29, 2010
    …orrelate function depend on the data dtype.
  2. DEP: remove deprecated econ kw from linalg.qr.

    rgommers committed Nov 29, 2010
    The mode keyword is now the same as that for numpy's qr. Closes #243.
  3. ENH: update mstats to use new zmap/zscore, remove z/zs. Closes #1195.

    rgommers committed Nov 29, 2010
    The plain stats functions zmap and zscore now also handle masked arrays, and
    they support more functionality that the versions that were removed.
  4. ENH: signal: add the 'nyq' keyword argument to firwin(), to be consis…

    warren.weckesser committed Nov 29, 2010
    …tent with the API of firwin2().
Commits on Nov 28, 2010
  1. ENH: signal: added the function firwin2 for constructing FIR filters …

    warren.weckesser committed Nov 28, 2010
    …with arbitrary frequency response.
  2. ENH: special/lpmv: implement degree down-recursion in associated lege…

    pv committed Nov 28, 2010
    …ndre polynomial (fixes #1296)
    
    Also implement support for negative orders and degrees, except in the
    cases V+M+1 <= 0, M < 0.
  3. BUG: special: fix mistaken divergence test in hyp2f1 for cases where …

    pv committed Nov 28, 2010
    …the series terminates (fixes #1298)