summaryrefslogtreecommitdiff
path: root/numpy/linalg/linalg.py
Commit message (Collapse)AuthorAgeFilesLines
* Keep singular values and eigenvalues as 1D arrays until the matrix indexingCharles Harris2008-04-291-7/+7
| | | | | controversy is settled. This will also keep code that uses diag(ev), from breaking.
* Fix test of lstsqr to work with matrix tests.Charles Harris2008-04-271-4/+4
| | | | Fix lstsq
* Add tests for matrix return types.Charles Harris2008-04-271-3/+4
| | | | | Fix cond computations for matrices. lstsq is currently broken for matrices, will fix shortly.
* Make functions in linalg.py accept nestes lists.Charles Harris2008-04-271-36/+55
| | | | | Use wrap to keep matrix environment intact. Base patch from nmb.
* ran reindent in preparation for the 1.1 releaseJarrod Millman2008-04-201-10/+10
|
* Added function for computing condition number, with tests and docs; closes #622.aarchiba2008-04-081-0/+38
|
* Factor out matrix_multiply from defmatrix. Based on a patch byStefan van der Walt2008-04-061-1/+5
| | | | Anne Archibald.
* ran reindentJarrod Millman2008-02-081-47/+47
|
* more docstring improvements from Pauli VirtanenJarrod Millman2007-12-291-12/+30
|
* docstring improvements from Pauli VirtanenJarrod Millman2007-12-291-252/+478
|
* Apply patch to fix ticket #557 (pinv causing error with empty arrays)Travis Oliphant2007-09-211-1/+8
|
* Add documentation for eigvals, eigvalsh, eig, and eigh.Charles Harris2007-05-131-4/+165
|
* whitespace cleanupTim Leslie2007-01-091-57/+59
|
* clean up unused imports and bad whitespaceTim Leslie2007-01-091-3/+3
|
* Fix hanging eig computation when inf or nan values are in the array. This ↵Travis Oliphant2006-12-061-1/+8
| | | | check is already in SciPy. We need it in NumPy too for some LaPack calls.
* Fix problem with dgesdd where the optimal work size is not computed correctly.Travis Oliphant2006-11-021-9/+0
|
* Fix qr decomposition. The dimensions m,n were taken from the fortran ordered ↵Charles Harris2006-10-161-21/+13
| | | | | | array instead of from the numpy ordered array and consequently were reversed.
* Fix-up tensor solve and tensor inv and rename to match tensordot.Travis Oliphant2006-10-131-25/+34
|
* Add a rcond parameter to the polyfit function and give it the double precisionCharles Harris2006-10-131-26/+26
| | | | | | | | | default value that dgelsd uses (rcondx=-1) on the principle of least surprise. Values of rcond less than this can also be useful, so a warning message and a bit of explanation was added to the documentation. The default value used by lstsq was set to the default (-1), and rcond in pinv has a default value of 1e-15.
* Add solvetensor and invtensorTravis Oliphant2006-10-131-2/+70
|
* Clean-up some un-needed default axes. Fix default axes of ma.sum and ma.productTravis Oliphant2006-08-291-1/+1
|
* Add axis arguments to various functions so as not to rely on the defaults.Travis Oliphant2006-08-291-4/+4
|
* Check for return value in linalg.detTravis Oliphant2006-08-141-0/+5
|
* Add qr decomposition to linalgTravis Oliphant2006-08-101-0/+90
|
* Move location of flatnonzeroTravis Oliphant2006-08-051-2/+2
|
* Fix uses of nonzero and add flatnonzeroTravis Oliphant2006-08-051-1/+1
|
* Revert change to lworkTravis Oliphant2006-07-311-1/+1
|
* Revert change and add better test for umathTravis Oliphant2006-07-311-1/+1
|
* Fix-up invalid casting back to real when complex result.Travis Oliphant2006-07-311-0/+11
|
* Fix missing return.Travis Oliphant2006-07-251-0/+1
|
* Add broadcasting behavior to random-number generators. Fix cholesky to keep ↵Travis Oliphant2006-07-251-1/+2
| | | | matrix return.
* Fix ticket #178 which was an error whenever multiple buffers needed to be ↵Travis Oliphant2006-07-181-4/+4
| | | | used to cast.
* numpy.linalg: fix bug where complex arrays weren't being returned.cookedm2006-07-141-3/+4
| | | | Also improved test cases.
* linalg routines will try to return their results as the same type as the ↵cookedm2006-07-141-158/+131
| | | | | | | | arguments. i.e., if you pass in a single or csingle array, that's what you get back. (Routines still use double-precision, though). Also some cleanups.
* squash remaining uses of typecodes in linalgcookedm2006-07-131-30/+30
|
* fix bug in eig with complex solutionscookedm2006-07-131-1/+1
|
* Fix error in linalg.pyTravis Oliphant2006-07-061-1/+4
|
* Fix buffered reduce, accumulate, and reduceat.Travis Oliphant2006-07-061-1/+1
|
* Convert linalg to use dtypes instead of typecodescookedm2006-07-041-36/+43
|
* Add some documentation to linalg.pyTravis Oliphant2006-06-271-0/+27
|
* Remove deprecated names from ma and tests.Travis Oliphant2006-06-021-1/+1
|
* Add 'q' to _array_kind and fix argmax which doesn't need to swap back axes.Travis Oliphant2006-05-161-2/+2
|
* change use of deprecated NewAxis to newaxisTim Leslie2006-03-181-3/+3
|
* Fix-up name-spaces.Travis Oliphant2006-03-151-1/+0
|
* Fix-up name-spaces.Travis Oliphant2006-03-151-5/+7
|
* fix imports so that all tests passTim Leslie2006-03-151-0/+1
|
* Isolate Numeric compatibility to numpy.linalg.oldTravis Oliphant2006-03-151-49/+20
|
* Committed much of ticket #36Travis Oliphant2006-03-151-21/+106
|
* Changed the C-API by making the a.flags object a builtin object for speed. ↵Travis Oliphant2006-03-151-1/+3
| | | | Also fixed issues with linalg and matrices.
* Run reindent.py (script distributed with Python) over the source to remove ↵cookedm2006-03-101-6/+4
| | | | extraneous whitespace
View Lorry Controller Status