Age | Commit message (Collapse) | Author | |
---|---|---|---|
2018-09-24 | implement for loop | Rémi Flamary | |
2018-09-24 | remove unused variable | Rémi Flamary | |
2018-09-24 | add contributor | Rémi Flamary | |
2018-09-24 | remove @ for python compatibility+ comments alexandre | Rémi Flamary | |
2018-09-24 | correct if error bug | Rémi Flamary | |
2018-09-24 | adding greenkhorn | alain | |
2018-09-07 | whitetrail pep8 | Nicolas Courty | |
2018-09-07 | stabThr and pep8 | Nicolas Courty | |
2018-09-07 | pep8 fixed (contd) | Nicolas Courty | |
2018-09-07 | pep8 normalization | Nicolas Courty | |
2018-09-07 | Wasserstein convolutional barycenter | Nicolas Courty | |
2018-07-24 | ensum tets marginals sinkhorn | Rémi Flamary | |
2018-07-24 | final makefile bench | Rémi Flamary | |
2018-07-24 | cancel einsum | Rémi Flamary | |
2018-07-24 | pb index | Rémi Flamary | |
2018-07-24 | correction size | Rémi Flamary | |
2018-07-24 | test eisum instead of dot | Rémi Flamary | |
2018-07-24 | pep8 all the way | Rémi Flamary | |
2018-07-24 | speedup einsum constraint violation | Rémi Flamary | |
2018-07-16 | Add comment & fix flake8 error | LeoGautheron | |
2018-07-11 | Speed-up Sinkhorn | LeoGautheron | |
Speed-up in 3 places: - the computation of pairwise distance is faster with sklearn.metrics.pairwise.euclidean_distances - faster computation of K = np.exp(-M / reg) - faster computation of the error every 10 iterations Example with this little script: import time import numpy as np import ot rng = np.random.RandomState(0) transport = ot.da.SinkhornTransport() time1 = time.time() Xs, ys, Xt = rng.randn(10000, 100), rng.randint(0, 2, size=10000), rng.randn(10000, 100) transport.fit(Xs=Xs, Xt=Xt) time2 = time.time() print("OT Computation Time {:6.2f} sec".format(time2-time1)) transport = ot.da.SinkhornLpl1Transport() transport.fit(Xs=Xs, ys=ys, Xt=Xt) time3 = time.time() print("OT LpL1 Computation Time {:6.2f} sec".format(time3-time2)) Before OT Computation Time 19.93 sec OT LpL1 Computation Time 133.43 sec After OT Computation Time 7.55 sec OT LpL1 Computation Time 82.25 sec | |||
2018-05-29 | update documentation for barycenter function | Rémi Flamary | |
2018-05-11 | pep8 | Rémi Flamary | |
2018-05-11 | update documentation in bregman | Rémi Flamary | |
2018-03-21 | add linear mapping test + autopep8 | Rémi Flamary | |
2017-07-26 | valid flake8 | Rémi Flamary | |
2017-07-26 | pep8 | Rémi Flamary | |
2017-07-26 | add license and authors on all modules | Rémi Flamary | |
2017-07-20 | pep8 | Alexandre Gramfort | |
2017-07-07 | add sinkhorbn2 +v3 | Rémi Flamary | |
2017-07-03 | debug bregman stabilized | Rémi Flamary | |
2017-06-20 | add clean zeros function for sparse distributions | Rémi Flamary | |
2017-06-13 | implement paralell sinkhorn stabilized | Rémi Flamary | |
2017-06-13 | implement paralell sinkhorn | Rémi Flamary | |
2017-06-13 | make sinkhorn more general with method selection | Rémi Flamary | |
2017-06-09 | add doc and correct encoding | Rémi Flamary | |
2017-04-20 | sinkhorn GPU implementation | Leo gautheron | |
2017-04-18 | little cleanup sinkhorn | Leo gautheron | |
2017-04-18 | Performance improvement sinkhorn | Leo gautheron | |
Doing the computation this way is equivalent and allows to reduce the space complexity required from O(max(a, b)^2) to O(a*b) (especially usefull to transport a small number of sources example to a lot of target) This also allows to decrease the computation time. | |||
2017-04-13 | update bregman file | Leo gautheron | |
- change commented prints to python3 compatible syntax - Correct issue that could cause the sinkhorn algo to stop with u and v containing nan/infinite numbers: - Assign uprev and vprev before changing u and v. - Then update u and v. - Then check if u and v contain nan, but ALSO infinite values. - if there are issues, then display error (with 2 r, not 3 :p) along with the iteration number (there may have errors at iteration 0) | |||
2017-02-07 | log return for sinkhorn | Rémi Flamary | |
2017-01-05 | more efficient sinkhorn | Rémi Flamary | |
2016-11-08 | travis CI | Rémi Flamary | |
2016-11-07 | v0.1.11 | Rémi Flamary | |
2016-11-07 | add log and epsilon scaling stabilizations | Rémi Flamary | |
2016-11-04 | doc nonlinear mapping estimation | Rémi Flamary | |
2016-10-28 | doc da.py | Rémi Flamary | |
2016-10-28 | doc datasets.py | Rémi Flamary | |
2016-10-28 | bregman doc finished | Rémi Flamary | |
2016-10-28 | update bregman with doc | Rémi Flamary | |