Dongryeol Lee
d8ef8d2ee4
Debuggin in process
2008-02-12 01:03:49 +00:00
Dongryeol Lee
1eee6feaac
Fixed the compilation error
2008-02-12 00:54:53 +00:00
Dongryeol Lee
63aaf1d852
Fixed the compilation error
2008-02-12 00:46:50 +00:00
Dongryeol Lee
aa010ab75a
Finished implementing the dense formulation of LPR
2008-02-12 00:38:00 +00:00
vasiloglou
59b86b4b0c
it seems to work
...
comments
The armijo rule is really slow
The statistical update of sigmas is good
We need to stay sometime in low sigmas otherwise we do not optimize
the trace
Next setp is BFGS method
2008-02-11 22:51:14 +00:00
Dongryeol Lee
f13a95e825
More fixes
2008-02-11 22:03:19 +00:00
Dongryeol Lee
5e3a004dc9
Algorithm reformulation
2008-02-11 21:35:37 +00:00
vasiloglou
38a107040e
it seems to me it is working, it is just a matter of tuning the parameters
2008-02-11 05:16:33 +00:00
Dongryeol Lee
259264690c
Finished the query tree initialization phase for dense version of LPR
2008-02-11 00:56:54 +00:00
Dongryeol Lee
68cfc553c0
Base case finished
2008-02-11 00:30:09 +00:00
Dongryeol Lee
ae5bb271ab
Added utility class for matrix
2008-02-10 22:46:29 +00:00
Dongryeol Lee
a8ccafa099
Added the implementation file for the dense version of LPR
2008-02-10 21:58:43 +00:00
Dongryeol Lee
30080ce695
Restructuring the utility class
2008-02-10 19:29:36 +00:00
Dongryeol Lee
2c3479c2b2
initial check in of the dense version of local polynomial regression
2008-02-10 19:19:50 +00:00
Dongryeol Lee
47fd5fa9ad
Renamed the naive algorithm to handle locally constant regression
2008-02-09 21:23:50 +00:00
vasiloglou
85aa1036b9
I had forgotten to add the test
2008-02-09 19:18:39 +00:00
vasiloglou
151a985de7
seems to run although it needs to be cross tested with a swiss roll example to
...
verify it. We are a little bit worried that the nearest neighbors might be wrong
and we have to check on that, because the optimum solution is 9% off the deasible
set. Need also to inser the constraint that the data are centered around zero.
2008-02-09 19:17:09 +00:00
vasiloglou
d55165fbb5
major revision
...
added armijo's method
it is not running properly yet
2008-02-09 01:00:57 +00:00
Dongryeol Lee
3be8cfc3b5
Krylov subspace version of local linear regression completed
2008-02-08 03:37:43 +00:00
Dongryeol Lee
8ec624c1cf
Performance enhancement
2008-02-08 03:21:36 +00:00
Dongryeol Lee
7d3d55911d
Took out the test routine
2008-02-08 02:19:04 +00:00
Dongryeol Lee
f1c1bcd3d6
Now phase 2 does relative error
2008-02-08 02:02:29 +00:00
Dongryeol Lee
e98c520087
Need to replace SYMMLQ method with Lanczos tridiagonalizer
2008-02-07 21:35:21 +00:00
Dongryeol Lee
45bb1686d8
Fixed a small performance bug
2008-02-07 20:11:14 +00:00
Dongryeol Lee
c9d020140f
Fixed a crucial bug in naive local linear
2008-02-07 05:40:02 +00:00
Dongryeol Lee
204b3a7590
Fixed the local linear naive code
2008-02-07 04:55:23 +00:00
Dongryeol Lee
add5522347
Fixed the local linear naive code
2008-02-07 04:55:01 +00:00
Dongryeol Lee
0d7dc02d09
Added the naive method
2008-02-07 04:32:27 +00:00
Dongryeol Lee
2c383c3b15
Added a new scaling function
2008-02-07 04:01:00 +00:00
Dongryeol Lee
348e0a9bbd
Changed the scaling
2008-02-07 04:00:13 +00:00
Dongryeol Lee
c1bc1ef321
Completely debugged locally linear regression
2008-02-07 03:28:41 +00:00
Dongryeol Lee
1e5a974fdf
The first working version of the matrix-free Krylov subspace local linear regression
2008-02-07 02:09:21 +00:00
Dongryeol Lee
9d766c7500
Added the test file for phase 1
2008-02-07 01:10:31 +00:00
Dongryeol Lee
f5fee10ff9
Phase 1 debugged
2008-02-07 00:47:53 +00:00
Dongryeol Lee
feaa56e449
Middle of debugging
2008-02-06 20:28:37 +00:00
Dongryeol Lee
8102757ad4
Currently debugging
2008-02-06 19:52:17 +00:00
Dongryeol Lee
36769a2bc5
SYMMLQ function finished. Debuggin in process
2008-02-06 19:08:45 +00:00
Dongryeol Lee
bc24336ee6
Fixed the postprocessing function for the 2nd phase
2008-02-06 03:12:22 +00:00
Dongryeol Lee
860957ebe7
Finished the rough structure except for the main Krylov subspace solver
2008-02-06 02:44:52 +00:00
Dongryeol Lee
80913b0d1e
Base case for the second phase completed...
2008-02-05 21:57:34 +00:00
Dongryeol Lee
892129d716
Ready to start the skeleton dual-tree code for the 2nd phase
2008-02-05 20:53:16 +00:00
vasiloglou
ed2383f4d9
it is running but the optimization error is increasing instead of decreasing
2008-02-05 20:12:34 +00:00
vasiloglou
ad11dbe66a
compiled
2008-02-05 19:29:51 +00:00
Dongryeol Lee
ed983bbeeb
More initialization routines completed
2008-02-05 19:22:15 +00:00
vasiloglou
e2f98f309b
minor edit on the build file
2008-02-05 18:32:28 +00:00
Dongryeol Lee
48edb52e2e
more structuring
2008-02-04 21:32:56 +00:00
Dongryeol Lee
09c0b23ecb
Basically ready to start the Krylov phase of the algorithm
2008-02-04 21:00:43 +00:00
Dongryeol Lee
fd3db6b92f
Changed the field variable names of Stat class to be more general
2008-02-04 19:47:40 +00:00
Dongryeol Lee
c7f8530eaa
Restructuring of the files
2008-02-04 19:23:45 +00:00
Dongryeol Lee
0c36d52282
Finished the postprocessing for the first phase of the algorithm
2008-02-04 16:31:03 +00:00
Dongryeol Lee