[MRG] Fix Gradient scaling in Partial GW solver (#602)

* new file:   ot/partial_gw.py

* remove partial_gw.py to update existing file partial.py

* fix pep8

---------

Co-authored-by: Rémi Flamary <[email protected]>
Co-authored-by: Cédric Vincent-Cuaz <[email protected]>

Author: 3 people

RELEASES.md

@@ -19,6 +19,7 @@
 - Fix same sign error for sr(F)GW conditional gradient solvers (PR #611)
 - Split `test/test_gromov.py` into `test/gromov/` (PR #619)
 - Fix (F)GW barycenter functions to support computing barycenter on 1 input + deprecate structures as lists (PR #628)
+- Fix line-search in partial GW and change default init to the interior of partial transport plans (PR #602)
 
 ## 0.9.3
 *January 2024*

ot/partial.py

@@ -4,12 +4,15 @@
 """
 
 # Author: Laetitia Chapel <[email protected]>
-# License: MIT License
+#             Yikun Bai < [email protected] >
+#             Cédric Vincent-Cuaz <[email protected]>
 
-import numpy as np
-from .lp import emd
-from .backend import get_backend
 from .utils import list_to_array
+from .backend import get_backend
+from .lp import emd
+import numpy as np
+
+# License: MIT License
 
 
 def partial_wasserstein_lagrange(a, b, M, reg_m=None, nb_dummies=1, log=False,
@@ -581,7 +584,7 @@ def partial_gromov_wasserstein(C1, C2, p, q, m=None, nb_dummies=1, G0=None,
                          " equal than min(|a|_1, |b|_1).")
 
     if G0 is None:
-        G0 = np.outer(p, q)
+        G0 = np.outer(p, q) * m / (np.sum(p) * np.sum(q))  # make sure |G0|=m, G01_m\leq p, G0.T1_n\leq q.
 
     dim_G_extended = (len(p) + nb_dummies, len(q) + nb_dummies)
     q_extended = np.append(q, [(np.sum(p) - m) / nb_dummies] * nb_dummies)
@@ -597,7 +600,7 @@ def partial_gromov_wasserstein(C1, C2, p, q, m=None, nb_dummies=1, G0=None,
 
         Gprev = np.copy(G0)
 
-        M = gwgrad_partial(C1, C2, G0)
+        M = 0.5 * gwgrad_partial(C1, C2, G0)  # rescaling the gradient with 0.5 for line-search while not changing Gc
         M_emd = np.zeros(dim_G_extended)
         M_emd[:len(p), :len(q)] = M
         M_emd[-nb_dummies:, -nb_dummies:] = np.max(M) * 1e2