Commit Graph

3716 Commits

Author SHA1 Message Date
Davis King
5ff051c53d Added negative_dot_product_distance and also clarified spec regarding thread
safety.
2013-01-27 16:05:25 -05:00
Davis King
dc9527beb2 Added cosine_distance. 2013-01-27 13:13:49 -05:00
Davis King
b5f77a0cc0 updated docs 2013-01-27 12:09:17 -05:00
Davis King
6aec0e08cf Added missing asserts 2013-01-27 11:59:47 -05:00
Davis King
364fd4969b Simplified the code a little by replacing some stuff with calls to
parallel_for_blocked()
2013-01-27 11:26:37 -05:00
Davis King
aeaf6e8798 updated docs 2013-01-27 11:15:28 -05:00
Davis King
446dcb43f0 Added more notes to spec 2013-01-27 11:11:04 -05:00
Davis King
da52cbb8b1 Added parallel_for() and parallel_for_blocked(). 2013-01-27 10:55:18 -05:00
Davis King
d2625a0dea made test more robust 2013-01-26 20:20:52 -05:00
Davis King
6e06b0bdf5 Made the distributed structural svm tools use the same improved job/buffering
rules as I recently added to the structural_svm_problem_threaded object.
2013-01-26 17:49:43 -05:00
Davis King
4f411d5a44 Made svm_struct_controller_node support network_address objects. 2013-01-26 17:25:12 -05:00
Davis King
cbc469bf42 Added comparison operators for network_address. 2013-01-26 17:23:39 -05:00
Davis King
596201e72d clarified print statements 2013-01-26 16:44:46 -05:00
Davis King
9d57628192 updated docs 2013-01-26 16:41:13 -05:00
Davis King
cd8ba14ff0 Added an example showing how to use second derivative information
when using the optimization tools.
2013-01-26 16:37:38 -05:00
Davis King
c4a63d77a1 Fixed bug in the axpy/scpy BLAS bindings which triggered when empty matrices
were assigned to each other.
2013-01-26 11:24:42 -05:00
Davis King
8d217377d7 Fixed a bug introduced yesterday 2013-01-26 10:28:44 -05:00
Davis King
e8d69f8958 Made tests more robust 2013-01-26 01:52:00 -05:00
Davis King
0487e37ed6 Changed code to avoid unnecessary cutting plane copying. This makes the
optimizer noticeably faster in some cases.
2013-01-26 01:51:28 -05:00
Davis King
c9bd5309ef Added missing const keywords 2013-01-26 01:01:47 -05:00
Davis King
633a9cdc9a Modified structural_svm_problem_threaded to reduce the amount of copying
overhead.
2013-01-26 00:44:58 -05:00
Davis King
51666563e3 Added BLAS bindings for xscal and xaxpy routines. 2013-01-26 00:20:03 -05:00
Davis King
c19bb9f36c Added another folder to blas search paths 2013-01-23 20:08:29 -05:00
Davis King
b97a64ed7f Changed the scripts so that the environment tests for makedocs only check
things really needed by makedocs.  Things needed by makerel are now in their
own check.  Also switched to unix2dos instead of unsing flip since unix2dos is
more common.

--HG--
rename : docs/testenv => docs/testenv_rel
2013-01-23 19:42:06 -05:00
Davis King
4eb8c5388b Changed the code so that each worker thread gets more things to do for each
job.  This should give better overall CPU utilization than the previous smaller
job sizes.
2013-01-23 18:49:31 -05:00
Davis King
14cea9da7d switched testenv to check for hg instead of svn 2013-01-23 18:39:12 -05:00
Davis King
b34b25a193 Expanded the set of folder we search when looking for BLAS and
LAPACK.
2013-01-22 18:57:22 -05:00
Davis King
98628bfea7 clarified spec 2013-01-21 18:02:05 -05:00
Davis King
014aa448a5 updated docs 2013-01-21 17:13:05 -05:00
Davis King
e341600ffb Added count_bits() and hamming_distance() 2013-01-21 17:06:27 -05:00
Davis King
2f61fe09c1 merged 2013-01-20 23:51:21 -05:00
Davis King
d707528209 Changed general_hash so it uses dlib::hash to hash integers rather
than returning them unmodified.
2013-01-20 23:50:44 -05:00
Davis King
a07590a7ad made unit test more robust 2013-01-20 14:21:10 -05:00
Davis King
ed9bd6376e clarified spec 2013-01-20 14:01:43 -05:00
Davis King
2b7b216f7c Added more cca() unit tests. 2013-01-20 13:43:50 -05:00
Davis King
a99fc5661a Made svd_fast() accept a wider range of matrices as arguments. 2013-01-20 11:50:01 -05:00
Davis King
6391a03440 Made some of this code a little more general. 2013-01-19 00:35:43 -05:00
Davis King
04f8d80361 Increased the default number of power iterations done by svd_fast()
from 0 to 1.
2013-01-19 00:35:19 -05:00
Davis King
9f9a720a23 updated docs 2013-01-19 00:17:16 -05:00
Davis King
b51d11484e Made the svd_fast() code a little more readable and memory efficient.
Also added the orthogonalize() function.
2013-01-19 00:15:29 -05:00
Davis King
246a60c249 Added a version of get_q() to qr_decomposition that allows the user
to get the Q matrix by reference rather than by value.
2013-01-18 23:17:56 -05:00
Davis King
9376502ce7 updated docs 2013-01-18 22:11:35 -05:00
Davis King
170cffad97 Added make_sparse_vector_inplace() 2013-01-18 22:10:42 -05:00
Davis King
c8b59522e1 refined spec a little 2013-01-17 20:43:16 -05:00
Davis King
d3c2655126 updated docs 2013-01-17 20:33:57 -05:00
Davis King
36bf764acc renamed function 2013-01-17 20:30:04 -05:00
Davis King
a2ae46cd03 Refactored the Platt scaling code a little. Now there is a function,
learn_platt_scaling(), that allows you to directly call the Platt scaling code
without supplying a trainer object.
2013-01-17 20:22:44 -05:00
Davis King
03ec260cb3 reformatted comments. 2013-01-17 19:20:38 -05:00
Davis King
91e8594b23 made examples slightly clearer 2013-01-17 19:18:35 -05:00
Davis King
d19048000c fixed a typo 2013-01-16 20:59:27 -05:00