diff options
author | RĂ©mi Flamary <remi.flamary@gmail.com> | 2021-10-27 08:41:08 +0200 |
---|---|---|
committer | GitHub <noreply@github.com> | 2021-10-27 08:41:08 +0200 |
commit | d7554331fc409fea48ee758fd630909dd9dc4827 (patch) | |
tree | 9b8ed4bf94c12d034d5fb1de5b7b5b76c23b4d05 /docs/source/quickstart.rst | |
parent | 76450dddf8dd62b9714b72e99ae075516246d433 (diff) |
[WIP] Sinkhorn in log space (#290)
* adda sinkhorn log and working sinkhorn2 function
* more tests pass
* more tests pass
* it works but not by default yet
* remove warningd
* update circleci doc
* update circleci doc
* new sinkhorn implemeted but not by default
* better
* doctest pass
* test doctest
* new test utils
* remove pep8 errors
* remove pep8 errors
* doc new implementtaion with log
* test sinkhorn 2
* doc for log implementation
Diffstat (limited to 'docs/source/quickstart.rst')
-rw-r--r-- | docs/source/quickstart.rst | 10 |
1 files changed, 9 insertions, 1 deletions
diff --git a/docs/source/quickstart.rst b/docs/source/quickstart.rst index fd046a1..232df7b 100644 --- a/docs/source/quickstart.rst +++ b/docs/source/quickstart.rst @@ -358,6 +358,11 @@ More details about the algorithms used are given in the following note. + :code:`method='sinkhorn'` calls :any:`ot.bregman.sinkhorn_knopp` the classic algorithm [2]_. + + :code:`method='sinkhorn_log'` calls :any:`ot.bregman.sinkhorn_log` the + sinkhorn algorithm in log space [2]_ that is more stable but can be + slower in numpy since `logsumexp` is not implmemented in parallel. + It is the recommended solver for applications that requires + differentiability with a small number of iterations. + :code:`method='sinkhorn_stabilized'` calls :any:`ot.bregman.sinkhorn_stabilized` the log stabilized version of the algorithm [9]_. + :code:`method='sinkhorn_epsilon_scaling'` calls @@ -389,7 +394,10 @@ More details about the algorithms used are given in the following note. solutions. Note that the greedy version of the Sinkhorn :any:`ot.bregman.greenkhorn` can also lead to a speedup and the screening version of the Sinkhorn :any:`ot.bregman.screenkhorn` aim a providing a - fast approximation of the Sinkhorn problem. + fast approximation of the Sinkhorn problem. For use of GPU and gradient + computation with small number of iterations we strongly recommend the + :any:`ot.bregman.sinkhorn_log` solver that will no need to check for + numerical problems. |