Closed GoogleCodeExporter closed 9 years ago
[deleted comment]
Yes, there is a tiny difference of the singular values but my application
requires the MATLAB one which is what it should be from the theory. Any
suggestions?
EJML by extractDIag():
71.5611696859336600
41.1896906595225700
36.6704302554985160
36.0066163727442500
33.5218874295711400
17.9164805604597780
16.0147592153640870
8.0622577482985600
8.0622577482985560
8.0622577482985510
7.2550641757216170
4.5489470485754270
4.1924929110640880
1.4142135623730978
1.4142135623730963
1.4142135623730958
1.4142135623730950
1.4142135623730950
1.4142135623730950
1.4142135623730947
1.4142135623730947
1.4142135623730930
1.2624026508386340
1.1230861141335210
0.9381052747817946
0.4573016700202984
0.4144298804285476
0.0000000000000039
MATLAB computes:
71.561169685933493
41.189690659522519
36.670430255498459
36.006616372744155
33.521887429571024
17.916480560459728
16.014759215364055
8.062257748298553
8.062257748298551
8.062257748298547
7.255064175721619
4.548947048575421
4.192492911064086
1.414213562373096
1.414213562373095
1.414213562373095
1.414213562373095
1.414213562373095
1.414213562373095
1.414213562373094
1.414213562373094
1.414213562373093
1.262402650838633
1.123086114133520
0.938105274781794
0.457301670020297
0.414429880428547
0.000000000000002
Original comment by farley...@dynagrid.net
on 4 Nov 2014 at 9:16
This issue is most likely caused by a fixed threshold being used for singular
values. Matlab uses a threshold based on matrix size. The code has been
modified to replicate matlab's behavior since it is more reasonable.
See complex branch on 685cb55d51176650bc866b89cf358e8f74e9a9af
Original comment by peter.ab...@gmail.com
on 12 Nov 2014 at 6:28
Original issue reported on code.google.com by
farley...@dynagrid.net
on 4 Nov 2014 at 3:57