nojhan
|
012d81f7e0
|
clean old useless comments
|
2012-07-10 14:11:22 +02:00 |
|
nojhan
|
487a76c863
|
use types from the distribution
|
2012-07-10 14:09:57 +02:00 |
|
nojhan
|
b0cbdf41ba
|
correct vector initialization within Eigen implemetation; delete useless asserts
|
2012-07-10 14:09:10 +02:00 |
|
nojhan
|
f405973736
|
use Eigen defaults column major store order
|
2012-07-10 14:07:34 +02:00 |
|
nojhan
|
2c0638aa8f
|
correct template typedef for NormalMulti* classes
|
2012-07-10 12:04:25 +02:00 |
|
nojhan
|
c0be5c9700
|
no need of a eigen test
|
2012-07-10 12:02:40 +02:00 |
|
nojhan
|
f8bae61095
|
use row major everywhere with Eigen
|
2012-07-09 22:58:55 +02:00 |
|
nojhan
|
f0564c233e
|
test support for Eigen implementations
|
2012-07-09 22:37:40 +02:00 |
|
nojhan
|
661ef08e44
|
working multi-normal sampler with eigen
Diagonal matrix are intermediate type, implicit conversion to matrix is needed.
|
2012-07-09 22:36:41 +02:00 |
|
nojhan
|
f3e1562a14
|
add the Eigen library implementations of normal distributions computations
|
2012-07-09 18:47:35 +02:00 |
|
nojhan
|
3eefe9cd25
|
deactivate eda_sa for the moment
|
2012-07-09 18:46:48 +02:00 |
|
nojhan
|
766ac33c5a
|
conditional build scripts with boost xor eigen
|
2012-07-09 16:26:19 +02:00 |
|
nojhan
|
42be24628e
|
Merge branch 'master' into cmaes
|
2012-07-09 11:32:13 +02:00 |
|
nojhan
|
2d19ff4e6d
|
prepare CMA-ES from the EDA example
|
2012-07-09 11:31:34 +02:00 |
|
nojhan
|
da4a7489f3
|
bugfix build: build applications, correct prototype for EDA
|
2012-07-09 11:29:48 +02:00 |
|
Caner Candan
|
6051ccb274
|
edoNormalMulti.h: added missing license header
|
2012-06-14 01:50:16 +02:00 |
|
Caner Candan
|
e851cc119b
|
Merge branch 'master' of ssh://eodev.git.sourceforge.net/gitroot/eodev/eodev
|
2012-03-06 21:56:15 +01:00 |
|
nojhan
|
426c3584be
|
remove warning about unused variable 'size' in edoSampler
|
2012-02-10 10:01:15 +01:00 |
|
Caner Candan
|
55cbeb0ca1
|
Merge branch 'master' of ssh://eodev.git.sourceforge.net/gitroot/eodev/eodev
|
2012-01-19 02:35:18 +01:00 |
|
nojhan
|
10a5712594
|
replace the assert on empty index in the dispatcher by a warning
|
2012-01-12 17:47:55 +01:00 |
|
nojhan
|
d10325d1ad
|
bypass repairer calls if the index list is empty in the dispatcher
|
2012-01-12 16:49:50 +01:00 |
|
nojhan
|
ebf33d8177
|
bugfixes concerning the mew api
|
2011-12-23 16:50:51 +01:00 |
|
nojhan
|
4805f72a80
|
use the new cholesky api in edoSamplerNormalMulti
|
2011-12-23 16:31:35 +01:00 |
|
nojhan
|
fdf3549b1c
|
Merge branch 'master' of ssh://localhost:8479/gitroot/eodev/eodev
Conflicts:
edo/src/edoSamplerNormalMulti.h
|
2011-12-23 16:06:52 +01:00 |
|
nojhan
|
c4d2f8eaf9
|
cholesky classes are now in separate files
|
2011-12-23 16:05:05 +01:00 |
|
Caner Candan
|
d7cc58def3
|
Merge branch 'master' of ssh://eodev.git.sourceforge.net/gitroot/eodev/eodev
|
2011-12-14 19:52:46 +01:00 |
|
nojhan
|
406f2abbc8
|
beautifying code for Cholesky classes
|
2011-12-14 15:53:57 +01:00 |
|
nojhan
|
3b66f04fd6
|
refactoring Cholesky decomposition in several classes
|
2011-12-14 15:49:27 +01:00 |
|
nojhan
|
164a81a10c
|
output pretty formated matrix in debug mode in the multi-normal sampler
|
2011-11-17 13:41:32 +01:00 |
|
nojhan
|
e09bb5551b
|
assert that the decimals parameters of the round repairer is <= 1.0
|
2011-11-17 13:40:26 +01:00 |
|
nojhan
|
9decda0c6a
|
cholesky factorization with rounding to zero
|
2011-11-15 17:10:46 +01:00 |
|
nojhan
|
06100a6b57
|
test binary now computes average errors of the differents cholesky factorization methods over random covariance matrices
|
2011-11-13 21:40:39 +01:00 |
|
nojhan
|
17ce8a90e3
|
cholesky test can take matrix size and precision arguments
|
2011-11-12 23:53:23 +01:00 |
|
nojhan
|
fe2cebc0e8
|
BUGFIX: factorized matrix are not symetric, cholesky factorization should process different types for covariance and decomposition + better format output for cholesky test
|
2011-11-12 23:44:31 +01:00 |
|
nojhan
|
e0f691c148
|
the complete robust cholesky factorization -- bugfix in LDLT: return the factorization LD^1/2 instead of L, thus no needs of accessors on D
|
2011-11-12 20:13:18 +01:00 |
|
nojhan
|
b2b1a96423
|
working robust cholesky factorization, with test binary
|
2011-11-12 15:52:18 +01:00 |
|
nojhan
|
d5d8f87a90
|
more pertinent names for cholesky factorization functions
|
2011-11-12 14:24:54 +01:00 |
|
nojhan
|
b9b3c486f9
|
first attempt at implementing a robust cholesky
|
2011-11-10 15:56:59 +01:00 |
|
nojhan
|
ce7f59d408
|
pass the Cholesky method from the multi-normal sampler to the decomposition sub-class
|
2011-11-10 14:07:58 +01:00 |
|
nojhan
|
fb82958253
|
more refactorization of the multi-normal sampler, now you can choose among two Cholesky factorization methods
|
2011-11-10 11:48:14 +01:00 |
|
nojhan
|
9fe6995df1
|
refactoring of the cholesky decomposition
|
2011-11-09 15:48:07 +01:00 |
|
nojhan
|
1af6715ddc
|
BUGFIX: should use the standard deviation when sampling a mono-normal law, because it have the same scale than the mean
|
2011-11-09 11:38:40 +01:00 |
|
nojhan
|
96dc505f92
|
bugfix: invalidate the individual after having repaired it
|
2011-11-09 10:35:41 +01:00 |
|
nojhan
|
391c9e9e22
|
use decimals instead of tens
|
2011-11-09 10:15:21 +01:00 |
|
nojhan
|
0e9d9f8c81
|
added a symetric round repairer around 0.5, plus a rounding at a given precision
|
2011-11-08 18:40:28 +01:00 |
|
nojhan
|
4d45970767
|
refactored vectors repairers that are using a function in a more generic way, added a binary function wrapper for vectors repairers, with the modulus function as an example
|
2011-11-08 17:55:01 +01:00 |
|
nojhan
|
979f33b721
|
ditaa text diagram explaining how the repairer dispatcher works
|
2011-11-07 15:44:36 +01:00 |
|
nojhan
|
03b69f1819
|
compilfix: correct template name for dispatcher's indexes
|
2011-11-03 11:13:32 +01:00 |
|
nojhan
|
da5b29146b
|
Merge branch 'master' of ssh://localhost:8479/gitroot/eodev/eodev
|
2011-11-03 10:58:27 +01:00 |
|
nojhan
|
d4c8a63edc
|
Merge branch 'master' of ssh://localhost:8479/gitroot/eodev/eodev
|
2011-11-03 10:55:53 +01:00 |
|