[WIP] Sinkhorn in log space

.circleci/config.yml

@@ -4,7 +4,7 @@ version: 2
 jobs:
     build_docs:
       docker:
-        - image: circleci/python:3.7-stretch
+        - image: cimg/python:3.9
       steps:
         - checkout
         - run:
@@ -34,18 +34,6 @@ jobs:
               - data-cache-0
               - pip-cache
 
-        - run:
-            name: Spin up Xvfb
-            command: |
-              /sbin/start-stop-daemon --start --quiet --pidfile /tmp/custom_xvfb_99.pid --make-pidfile --background --exec /usr/bin/Xvfb -- :99 -screen 0 1400x900x24 -ac +extension GLX +render -noreset;
-
-        # https://github.com/ContinuumIO/anaconda-issues/issues/9190#issuecomment-386508136
-        # https://github.com/golemfactory/golem/issues/1019
-        - run:
-            name: Fix libgcc_s.so.1 pthread_cancel bug
-            command: |
-              sudo apt-get install qt5-default
-
         - run:
             name: Get Python running
             command: |

README.md

@@ -20,7 +20,7 @@ POT provides the following generic OT solvers (links to examples):
 
 * [OT Network Simplex solver](https://pythonot.github.io/auto_examples/plot_OT_1D.html) for the linear program/ Earth Movers Distance [1] .
 * [Conditional gradient](https://pythonot.github.io/auto_examples/plot_optim_OTreg.html) [6] and [Generalized conditional gradient](https://pythonot.github.io/auto_examples/plot_optim_OTreg.html) for regularized OT [7].
-* Entropic regularization OT solver with [Sinkhorn Knopp Algorithm](https://pythonot.github.io/auto_examples/plot_OT_1D.html) [2] , stabilized version [9] [10], greedy Sinkhorn [22] and [Screening Sinkhorn [26] ](https://pythonot.github.io/auto_examples/plot_screenkhorn_1D.html).
+* Entropic regularization OT solver with [Sinkhorn Knopp Algorithm](https://pythonot.github.io/auto_examples/plot_OT_1D.html) [2] , stabilized version [9] [10] [34], greedy Sinkhorn [22] and [Screening Sinkhorn [26] ](https://pythonot.github.io/auto_examples/plot_screenkhorn_1D.html).
 * Bregman projections for [Wasserstein barycenter](https://pythonot.github.io/auto_examples/barycenters/plot_barycenter_lp_vs_entropic.html) [3], [convolutional barycenter](https://pythonot.github.io/auto_examples/barycenters/plot_convolutional_barycenter.html) [21]  and unmixing [4].
 * Sinkhorn divergence [23] and entropic regularization OT  from empirical data.
 * [Smooth optimal transport solvers](https://pythonot.github.io/auto_examples/plot_OT_1D_smooth.html) (dual and semi-dual) for KL and squared L2 regularizations [17].
@@ -290,3 +290,5 @@ You can also post bug reports and feature requests in Github issues. Make sure t
 [32] Huang, M., Ma S., Lai, L. (2021). [A Riemannian Block Coordinate Descent Method for Computing the Projection Robust Wasserstein Distance](http://proceedings.mlr.press/v139/huang21e.html), Proceedings of the 38th International Conference on Machine Learning (ICML).
 
 [33] Kerdoncuff T., Emonet R., Marc S. [Sampled Gromov Wasserstein](https://hal.archives-ouvertes.fr/hal-03232509/document), Machine Learning Journal (MJL), 2021
+
+[34] Feydy, J., Séjourné, T., Vialard, F. X., Amari, S. I., Trouvé, A., & Peyré, G. (2019, April). Interpolating between optimal transport and MMD using Sinkhorn divergences. In The 22nd International Conference on Artificial Intelligence and Statistics (pp. 2681-2690). PMLR.

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.