[WIP] Improved docs and changed scipy version (#163)

* Improved docs and changed scipy version

* Fixed dependency bug in setup.py

* dependencies set to minimal versions for tests

* add requirements file

* added minimal version build for scipy (testing 1.2)

* bugfix in minimal deps build

* (yet another) bugfix in minimal deps build

* minimal deps now reflect README.md

* minimal deps: no autograd nor pymanopt

* refactored workflow names

* minimal deps: no doctests

* minimal deps: numpy 1.16

* trigger GH Actions on PR

* better merge

* re-add minimal-deps...

* bugfix in yaml

* enforce np>=1.16

* enforce scipy and cython versions too

* requires / install_requires

* requires / install_requires / requires

* setup_requires

Co-authored-by: Rémi Flamary <remi.flamary@gmail.com>

15 files changed, 110 insertions, 88 deletions

diff --git a/.github/requirements_strict.txt b/.github/requirements_strict.txt
new file mode 100644
index 0000000..d7539c5
--- /dev/null
+++ b/.github/requirements_strict.txt
@@ -0,0 +1,7 @@
+numpy==1.16.*
+scipy==1.0.*
+cython==0.23.*
+matplotlib
+cvxopt
+scikit-learn
+pytest
diff --git a/.github/workflows/build_tests.yml b/.github/workflows/build_tests.yml
index 652655f..41b08b3 100644
--- a/.github/workflows/build_tests.yml
+++ b/.github/workflows/build_tests.yml
@@ -2,9 +2,9 @@ name: build
 
 on:
   push:
-    branches:
-      - '**'
+
   pull_request:
+
   create:
     branches:
       - 'master'
@@ -49,6 +49,34 @@ jobs:
         codecov
 
 
+  linux-minimal-deps:
+
+    runs-on: ubuntu-latest
+    strategy:
+      max-parallel: 4
+      matrix:
+        python-version: [3.6]
+
+    steps:
+    - uses: actions/checkout@v1
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v1
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install -r .github/requirements_strict.txt
+        pip install pytest
+        pip install -U "sklearn"
+    - name: Install POT
+      run: |
+        pip install -e .
+    - name: Run tests
+      run: |
+        python -m pytest -v test/ ot/ --ignore ot/gpu/
+
+
   macos:
      runs-on: macOS-latest
      strategy:
diff --git a/README.md b/README.md
index b9b8f45..5ee4cee 100644
--- a/README.md
+++ b/README.md
@@ -69,7 +69,7 @@ year={2017}
 
 The library has been tested on Linux, MacOSX and Windows. It requires a C++ compiler for building/installing the EMD solver and relies on the following Python modules:
 
-- Numpy (>=1.11)
+- Numpy (>=1.16)
 - Scipy (>=1.0)
 - Cython (>=0.23)
 - Matplotlib (>=1.5)
diff --git a/examples/barycenters/plot_free_support_barycenter.py b/examples/barycenters/plot_free_support_barycenter.py
index 64b89e4..27ddc8e 100644
--- a/examples/barycenters/plot_free_support_barycenter.py
+++ b/examples/barycenters/plot_free_support_barycenter.py
@@ -4,7 +4,7 @@
 2D free support Wasserstein barycenters of distributions
 ====================================================
 
-Illustration of 2D Wasserstein barycenters if discributions that are weighted
+Illustration of 2D Wasserstein barycenters if distributions are weighted
 sum of diracs.
 
 """
@@ -21,7 +21,7 @@ import ot
 ##############################################################################
 # Generate data
 # -------------
-#%% parameters and data generation
+
 N = 3
 d = 2
 measures_locations = []
@@ -46,7 +46,7 @@ for i in range(N):
 
 ##############################################################################
 # Compute free support barycenter
-# -------------
+# -------------------------------
 
 k = 10  # number of Diracs of the barycenter
 X_init = np.random.normal(0., 1., (k, d))  # initial Dirac locations
diff --git a/examples/domain-adaptation/plot_otda_d2.py b/examples/domain-adaptation/plot_otda_d2.py
index f49a570..d8b2a93 100644
--- a/examples/domain-adaptation/plot_otda_d2.py
+++ b/examples/domain-adaptation/plot_otda_d2.py
@@ -25,7 +25,7 @@ import ot
 import ot.plot
 
 ##############################################################################
-# generate data
+# Generate data
 # -------------
 
 n_samples_source = 150
diff --git a/examples/domain-adaptation/plot_otda_mapping.py b/examples/domain-adaptation/plot_otda_mapping.py
index ded2bdf..d21d3c9 100644
--- a/examples/domain-adaptation/plot_otda_mapping.py
+++ b/examples/domain-adaptation/plot_otda_mapping.py
@@ -9,8 +9,8 @@ time both the coupling transport and approximate the transport map with either
 a linear or a kernelized mapping as introduced in [8].
 
 [8] M. Perrot, N. Courty, R. Flamary, A. Habrard,
-    "Mapping estimation for discrete optimal transport",
-    Neural Information Processing Systems (NIPS), 2016.
+"Mapping estimation for discrete optimal transport",
+Neural Information Processing Systems (NIPS), 2016.
 """
 
 # Authors: Remi Flamary <remi.flamary@unice.fr>
diff --git a/examples/domain-adaptation/plot_otda_mapping_colors_images.py b/examples/domain-adaptation/plot_otda_mapping_colors_images.py
index 9d3a7c7..ee5c8b0 100644
--- a/examples/domain-adaptation/plot_otda_mapping_colors_images.py
+++ b/examples/domain-adaptation/plot_otda_mapping_colors_images.py
@@ -9,6 +9,7 @@ estimation [8].
 
 [6] Ferradans, S., Papadakis, N., Peyre, G., & Aujol, J. F. (2014). Regularized
 discrete optimal transport. SIAM Journal on Imaging Sciences, 7(3), 1853-1882.
+
 [8] M. Perrot, N. Courty, R. Flamary, A. Habrard, "Mapping estimation for
 discrete optimal transport", Neural Information Processing Systems (NIPS), 2016.
 
diff --git a/examples/gromov/plot_barycenter_fgw.py b/examples/gromov/plot_barycenter_fgw.py
index 77b0370..3f81765 100644
--- a/examples/gromov/plot_barycenter_fgw.py
+++ b/examples/gromov/plot_barycenter_fgw.py
@@ -4,14 +4,15 @@
 Plot graphs' barycenter using FGW
 =================================
 
-This example illustrates the computation barycenter of labeled graphs using FGW
+This example illustrates the computation barycenter of labeled graphs using
+FGW [18].
 
 Requires networkx >=2
 
-.. [18] Vayer Titouan, Chapel Laetitia, Flamary R{\'e}mi, Tavenard Romain
-      and Courty Nicolas
-    "Optimal Transport for structured data with application on graphs"
-    International Conference on Machine Learning (ICML). 2019.
+[18] Vayer Titouan, Chapel Laetitia, Flamary Rémi, Tavenard Romain
+and Courty Nicolas
+"Optimal Transport for structured data with application on graphs"
+International Conference on Machine Learning (ICML). 2019.
 
 """
 
diff --git a/examples/gromov/plot_fgw.py b/examples/gromov/plot_fgw.py
index 73e486e..97fe619 100644
--- a/examples/gromov/plot_fgw.py
+++ b/examples/gromov/plot_fgw.py
@@ -4,12 +4,12 @@
 Plot Fused-gromov-Wasserstein
 ==============================
 
-This example illustrates the computation of FGW for 1D measures[18].
+This example illustrates the computation of FGW for 1D measures [18].
 
-.. [18] Vayer Titouan, Chapel Laetitia, Flamary R{\'e}mi, Tavenard Romain
-      and Courty Nicolas
-    "Optimal Transport for structured data with application on graphs"
-    International Conference on Machine Learning (ICML). 2019.
+[18] Vayer Titouan, Chapel Laetitia, Flamary Rémi, Tavenard Romain
+and Courty Nicolas
+"Optimal Transport for structured data with application on graphs"
+International Conference on Machine Learning (ICML). 2019.
 
 """
 
diff --git a/examples/plot_compute_emd.py b/examples/plot_compute_emd.py
index 3340115..527a847 100644
--- a/examples/plot_compute_emd.py
+++ b/examples/plot_compute_emd.py
@@ -4,8 +4,8 @@
 Plot multiple EMD
 =================
 
-Shows how to compute multiple EMD and Sinkhorn with two differnt
-ground metrics and plot their values for diffeent distributions.
+Shows how to compute multiple EMD and Sinkhorn with two different
+ground metrics and plot their values for different distributions.
 
 
 """
diff --git a/examples/plot_optim_OTreg.py b/examples/plot_optim_OTreg.py
index 51e2fdc..5eb15bd 100644
--- a/examples/plot_optim_OTreg.py
+++ b/examples/plot_optim_OTreg.py
@@ -6,7 +6,7 @@ Regularized OT with generic solver
 
 Illustrates the use of the generic solver for regularized OT with
 user-designed regularization term. It uses Conditional gradient as in [6] and
-generalized Conditional Gradient as proposed in [5][7].
+generalized Conditional Gradient as proposed in [5,7].
 
 
 [5] N. Courty; R. Flamary; D. Tuia; A. Rakotomamonjy, Optimal Transport for
@@ -14,8 +14,8 @@ Domain Adaptation, in IEEE Transactions on Pattern Analysis and Machine
 Intelligence , vol.PP, no.99, pp.1-1.
 
 [6] Ferradans, S., Papadakis, N., Peyré, G., & Aujol, J. F. (2014).
-Regularized discrete optimal transport. SIAM Journal on Imaging Sciences,
-7(3), 1853-1882.
+Regularized discrete optimal transport. SIAM Journal on Imaging
+Sciences, 7(3), 1853-1882.
 
 [7] Rakotomamonjy, A., Flamary, R., & Courty, N. (2015). Generalized
 conditional gradient: analysis of convergence and applications.
diff --git a/examples/plot_screenkhorn_1D.py b/examples/plot_screenkhorn_1D.py
index 840ead8..785642a 100644
--- a/examples/plot_screenkhorn_1D.py
+++ b/examples/plot_screenkhorn_1D.py
@@ -4,8 +4,11 @@
 1D Screened optimal transport
 ===============================
 
-This example illustrates the computation of Screenkhorn:
-Screening Sinkhorn Algorithm for Optimal transport.
+This example illustrates the computation of Screenkhorn [26].
+
+[26] Alaya M. Z., Bérar M., Gasso G., Rakotomamonjy A. (2019).
+Screening Sinkhorn Algorithm for Regularized Optimal Transport,
+Advances in Neural Information Processing Systems 33 (NeurIPS).
 """
 
 # Author: Mokhtar Z. Alaya <mokhtarzahdi.alaya@gmail.com>
diff --git a/examples/plot_stochastic.py b/examples/plot_stochastic.py
index 742f8d9..3a1ef31 100644
--- a/examples/plot_stochastic.py
+++ b/examples/plot_stochastic.py
@@ -1,10 +1,18 @@
 """
-==========================
+===================
 Stochastic examples
-==========================
+===================
 
 This example is designed to show how to use the stochatic optimization
-algorithms for descrete and semicontinous measures from the POT library.
+algorithms for discrete and semi-continuous measures from the POT library.
+
+[18] Genevay, A., Cuturi, M., Peyré, G. & Bach, F.
+Stochastic Optimization for Large-scale Optimal Transport.
+Advances in Neural Information Processing Systems (2016).
+
+[19] Seguy, V., Bhushan Damodaran, B., Flamary, R., Courty, N., Rolet, A. &
+Blondel, M. Large-scale Optimal Transport and Mapping Estimation.
+International Conference on Learning Representation (2018)
 
 """
 
@@ -19,16 +27,14 @@ import ot.plot
 
 
 #############################################################################
-# COMPUTE TRANSPORTATION MATRIX FOR SEMI-DUAL PROBLEM
-#############################################################################
-#############################################################################
-# DISCRETE CASE:
+# Compute the Transportation Matrix for the Semi-Dual Problem
+# -----------------------------------------------------------
 #
-# Sample two discrete measures for the discrete case
-# ---------------------------------------------
+# Discrete case
+# `````````````
 #
-# Define 2 discrete measures a and b, the points where are defined the source
-# and the target measures and finally the cost matrix c.
+# Sample two discrete measures for the discrete case and compute their cost
+# matrix c.
 
 n_source = 7
 n_target = 4
@@ -44,12 +50,7 @@ Y_target = rng.randn(n_target, 2)
 M = ot.dist(X_source, Y_target)
 
 #############################################################################
-#
 # Call the "SAG" method to find the transportation matrix in the discrete case
-# ---------------------------------------------
-#
-# Define the method "SAG", call ot.solve_semi_dual_entropic and plot the
-# results.
 
 method = "SAG"
 sag_pi = ot.stochastic.solve_semi_dual_entropic(a, b, M, reg, method,
@@ -57,14 +58,12 @@ sag_pi = ot.stochastic.solve_semi_dual_entropic(a, b, M, reg, method,
 print(sag_pi)
 
 #############################################################################
-# SEMICONTINOUS CASE:
+# Semi-Continuous Case
+# ````````````````````
 #
 # Sample one general measure a, one discrete measures b for the semicontinous
-# case
-# ---------------------------------------------
-#
-# Define one general measure a, one discrete measures b, the points where
-# are defined the source and the target measures and finally the cost matrix c.
+# case, the points where source and target measures are defined and compute the
+# cost matrix.
 
 n_source = 7
 n_target = 4
@@ -81,13 +80,8 @@ Y_target = rng.randn(n_target, 2)
 M = ot.dist(X_source, Y_target)
 
 #############################################################################
-#
 # Call the "ASGD" method to find the transportation matrix in the semicontinous
-# case
-# ---------------------------------------------
-#
-# Define the method "ASGD", call ot.solve_semi_dual_entropic and plot the
-# results.
+# case.
 
 method = "ASGD"
 asgd_pi, log_asgd = ot.stochastic.solve_semi_dual_entropic(a, b, M, reg, method,
@@ -96,23 +90,17 @@ print(log_asgd['alpha'], log_asgd['beta'])
 print(asgd_pi)
 
 #############################################################################
-#
 # Compare the results with the Sinkhorn algorithm
-# ---------------------------------------------
-#
-# Call the Sinkhorn algorithm from POT
 
 sinkhorn_pi = ot.sinkhorn(a, b, M, reg)
 print(sinkhorn_pi)
 
 
 ##############################################################################
-# PLOT TRANSPORTATION MATRIX
-##############################################################################
-
-##############################################################################
-# Plot SAG results
-# ----------------
+# Plot Transportation Matrices
+# ````````````````````````````
+#
+# For SAG
 
 pl.figure(4, figsize=(5, 5))
 ot.plot.plot1D_mat(a, b, sag_pi, 'semi-dual : OT matrix SAG')
@@ -120,8 +108,7 @@ pl.show()
 
 
 ##############################################################################
-# Plot ASGD results
-# -----------------
+# For ASGD
 
 pl.figure(4, figsize=(5, 5))
 ot.plot.plot1D_mat(a, b, asgd_pi, 'semi-dual : OT matrix ASGD')
@@ -129,8 +116,7 @@ pl.show()
 
 
 ##############################################################################
-# Plot Sinkhorn results
-# ---------------------
+# For Sinkhorn
 
 pl.figure(4, figsize=(5, 5))
 ot.plot.plot1D_mat(a, b, sinkhorn_pi, 'OT matrix Sinkhorn')
@@ -138,17 +124,14 @@ pl.show()
 
 
 #############################################################################
-# COMPUTE TRANSPORTATION MATRIX FOR DUAL PROBLEM
-#############################################################################
-#############################################################################
-# SEMICONTINOUS CASE:
+# Compute the Transportation Matrix for the Dual Problem
+# ------------------------------------------------------
 #
-# Sample one general measure a, one discrete measures b for the semicontinous
-# case
-# ---------------------------------------------
+# Semi-continuous case
+# ````````````````````
 #
-# Define one general measure a, one discrete measures b, the points where
-# are defined the source and the target measures and finally the cost matrix c.
+# Sample one general measure a, one discrete measures b for the semi-continuous
+# case and compute the cost matrix c.
 
 n_source = 7
 n_target = 4
@@ -169,10 +152,7 @@ M = ot.dist(X_source, Y_target)
 #############################################################################
 #
 # Call the "SGD" dual method to find the transportation matrix in the
-# semicontinous case
-# ---------------------------------------------
-#
-# Call ot.solve_dual_entropic and plot the results.
+# semi-continuous case
 
 sgd_dual_pi, log_sgd = ot.stochastic.solve_dual_entropic(a, b, M, reg,
                                                          batch_size, numItermax,
@@ -183,7 +163,7 @@ print(sgd_dual_pi)
 #############################################################################
 #
 # Compare the results with the Sinkhorn algorithm
-# ---------------------------------------------
+# ```````````````````````````````````````````````
 #
 # Call the Sinkhorn algorithm from POT
 
@@ -191,8 +171,10 @@ sinkhorn_pi = ot.sinkhorn(a, b, M, reg)
 print(sinkhorn_pi)
 
 ##############################################################################
-# Plot  SGD results
-# -----------------
+# Plot Transportation Matrices
+# ````````````````````````````
+#
+# For SGD
 
 pl.figure(4, figsize=(5, 5))
 ot.plot.plot1D_mat(a, b, sgd_dual_pi, 'dual : OT matrix SGD')
@@ -200,8 +182,7 @@ pl.show()
 
 
 ##############################################################################
-# Plot Sinkhorn results
-# ---------------------
+# For Sinkhorn
 
 pl.figure(4, figsize=(5, 5))
 ot.plot.plot1D_mat(a, b, sinkhorn_pi, 'OT matrix Sinkhorn')
diff --git a/requirements.txt b/requirements.txt
index bee22f7..331dd57 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,5 +1,5 @@
 numpy
-scipy>=1.0
+scipy>=1.3
 cython
 matplotlib
 autograd
diff --git a/setup.py b/setup.py
index 4640d00..91c24d9 100755
--- a/setup.py
+++ b/setup.py
@@ -67,7 +67,8 @@ setup(name='POT',
       scripts=[],
       data_files=[],
       requires=["numpy", "scipy", "cython"],
-      install_requires=["numpy", "scipy", "cython"],
+      setup_requires=["numpy>=1.16", "scipy>=1.0", "cython>=0.23"],
+      install_requires=["numpy>=1.16", "scipy>=1.0", "cython>=0.23"],
       classifiers=[
           'Development Status :: 5 - Production/Stable',
           'Intended Audience :: Developers',