kortschak
c4b4903572
Remove unnecessary zeroing code
2014-01-10 12:08:22 +10:30
kortschak
25c207cb64
Change from JAMA name
2014-01-09 10:47:49 +10:30
kortschak
3b5083ad84
Remove nx as unnecessary
...
This clarifies the constraints we are working in.
2014-01-09 10:44:11 +10:30
kortschak
36eada56ab
Make it very clear what we are doing here.
2014-01-09 10:35:46 +10:30
btracey
7096590f28
Removed commented fmt
2014-01-08 15:45:42 -08:00
btracey
68efb7a653
Fixed QR solve code to give correct answer for skinny matrices
2014-01-08 15:45:42 -08:00
btracey
f1706a148f
Removed dead code in qr test
2014-01-08 15:45:42 -08:00
btracey
e88f486561
Working version of Solve
2014-01-08 15:45:42 -08:00
kortschak
430f7a3c1b
Improve QR comment
2014-01-09 09:42:32 +10:30
kortschak
f75c287c0f
Clean imports
2014-01-09 09:18:32 +10:30
kortschak
d07426118b
Enforce QRD shape restriction
2014-01-09 09:13:40 +10:30
btracey
7506701a44
Removed fat test, algorithm doesn't work in that case
2014-01-08 12:43:48 -08:00
btracey
3fb71a63c2
Added Fat QR test
2014-01-08 12:18:19 -08:00
btracey
f375e1ed84
Added skinny QR test
2014-01-08 12:17:26 -08:00
btracey
38651e3475
Added square test for QR factorization
2014-01-08 12:14:13 -08:00
kortschak
4308ceaa16
Fix illegal index panic
2014-01-08 16:00:35 +10:30
kortschak
70c87063ca
Improve panic test handling
2014-01-08 16:00:08 +10:30
kortschak
64b1087c7c
Fix QR solve shape error
2014-01-08 15:33:52 +10:30
kortschak
5823c3fdfb
Clean up tests
2014-01-08 15:21:47 +10:30
btracey
f5c64a4270
Fixed tests to have answers from a reliable source
2014-01-07 18:29:22 -08:00
btracey
022dec19df
Added more tests to Solve
2014-01-07 18:12:53 -08:00
btracey
56b7e0d9f1
Changed test to use flatten
2014-01-07 16:59:50 -08:00
btracey
78992998ad
Started adding tests for Solve
2014-01-07 16:46:07 -08:00
kortschak
b3019360f6
Fix dimensio typo
...
This caused a panic with b matrices that are not square.
2014-01-08 10:30:33 +10:30
kortschak
b10f3a00f3
Change NewDense signature and behaviour
...
This is an API breaking change.
NewDense now panics if len(mat) != r*c, unless mat == nil. When mat is
nil a new, correctly sized slice is allocated.
2014-01-08 09:56:39 +10:30
kortschak
fadcd6dd7f
Avoid confusion with the name realloc
...
Fixes issue 1.
2013-12-27 19:40:43 +10:30
dane-unltd
c3ee9fcffd
Fixed Dot
2013-11-20 19:48:39 +01:00
kortschak
172ec6bf55
Add test for Norm
...
Norm(±2) is, not surprisingly, difficult to pin down. The expected
values here agree with numpy well for Norm(2), less will with MATLAB
and the expectation for Norm(-2) is a fair amount (FWIW at this
magnitude) away from the numpy and MATLAB values which agree quite
well with each other.
2013-11-20 09:53:11 +10:30
kortschak
01d2d8774d
Fill in minimal Norm(±2) code
...
We should be able to handle wide matrices, but this will come later.
2013-11-20 09:12:51 +10:30
kortschak
8a6654a9a1
Merge mat64/la into mat64
...
This has been done to resolve a circular dependency issue that prevents
Norm(2) and Norm(-2) from being implemented.
2013-11-20 09:07:33 +10:30
kortschak
e7389856e5
Forbid loading column major blas matrix
...
This may change. Perhaps we can have an allocating LoadBlas for column
major blas matrices. Thinking.
2013-11-15 12:53:55 +10:30
kortschak
0e660dd08a
Allow client to see which engine is registered
2013-11-15 12:46:07 +10:30
kortschak
ffbeadd5fa
Remove intermediate vars required for conditional order
2013-11-15 12:44:27 +10:30
kortschak
2350c3a978
Remove order conditionals
2013-11-15 11:15:27 +10:30
kortschak
26b1734f85
Make order constant
...
This is a start to removing column major handling of matrices.
2013-11-15 10:50:23 +10:30
kortschak
dd2cd47b0c
Simplify L\U tests
2013-11-05 09:34:29 +10:30
kortschak
a9c60e674c
Fix off by one error in L() code
...
This was missed in tests because the in-place special case was not
tested. Thanks to David Neumann for the heads up.
2013-11-05 09:27:15 +10:30
kortschak
04796ca6f7
We are not going to implement Det or Inv
...
This functionality is provided in la.
2013-10-31 16:06:59 +10:30
kortschak
f36c5cf03b
Update error strings
2013-10-31 16:04:31 +10:30
kortschak
1a48607293
Add a check for blas engine registration
...
This will give a more meaningful message when people forget to do this.
2013-10-31 15:57:26 +10:30
kortschak
cb949b4564
Fix incorrect dimension adjustment for View
...
We were previously mixing semantics, with expectedly bad outcomes.
2013-10-31 13:30:54 +10:30
kortschak
6e75c43e8e
Add test for Inverse
2013-10-31 11:31:43 +10:30
kortschak
effc954a86
Add check for self determinant calculation
...
If the matrix can calculate its own determinant, let it.
2013-10-31 11:16:31 +10:30
kortschak
76156164b4
Wrap associated data and methods
...
The API was too cluttered. This should fix that.
2013-10-31 10:52:38 +10:30
kortschak
7b8dc30c92
Add DenseCopyOf convenience function
2013-10-31 10:50:10 +10:30
kortschak
79c029df34
Add convenience functions for inverse and solve
2013-10-31 09:54:45 +10:30
kortschak
b0fcacac77
LU decomposition
...
Test data is duplicated in source; this will be fixed after a helper is
added to mat64.Dense.
2013-10-31 09:29:19 +10:30
kortschak
8d32714967
Add Cholesky and QR decomposition
...
The API is starting to get a little cluttered. We should consider
wrapping results from decompositions in structs.
2013-10-30 12:15:25 +10:30
kortschak
fd73bc1b29
Add failing case matrix for fat matrix SVD
...
This passes as long as we only ask for sigma, and we allow sigma to be
one element longer than necessary.
2013-10-30 08:37:52 +10:30
kortschak
a2d1051f45
Remove unnecessary brackets
2013-10-29 22:06:29 +10:30
