diff options
Diffstat (limited to 'src/python')
| -rw-r--r-- | src/python/CMakeLists.txt | 11 | ||||
| -rw-r--r-- | src/python/doc/installation.rst | 9 | ||||
| -rw-r--r-- | src/python/doc/rips_complex_sum.inc | 22 | ||||
| -rw-r--r-- | src/python/doc/rips_complex_user.rst | 6 | ||||
| -rw-r--r-- | src/python/gudhi/simplex_tree.pxd | 1 | ||||
| -rw-r--r-- | src/python/gudhi/simplex_tree.pyx | 22 | ||||
| -rw-r--r-- | src/python/gudhi/subsampling.pyx | 21 | ||||
| -rwxr-xr-x | src/python/test/test_simplex_tree.py | 22 | ||||
| -rwxr-xr-x | src/python/test/test_wasserstein_distance.py | 24 | ||||
| -rwxr-xr-x | src/python/test/test_wasserstein_with_tensors.py | 47 |
10 files changed, 130 insertions, 55 deletions
diff --git a/src/python/CMakeLists.txt b/src/python/CMakeLists.txt index 4f26481e..c09996fe 100644 --- a/src/python/CMakeLists.txt +++ b/src/python/CMakeLists.txt @@ -103,6 +103,9 @@ if(PYTHONINTERP_FOUND) if(EAGERPY_FOUND) add_gudhi_debug_info("EagerPy version ${EAGERPY_VERSION}") endif() + if(TENSORFLOW_FOUND) + add_gudhi_debug_info("TensorFlow version ${TENSORFLOW_VERSION}") + endif() set(GUDHI_PYTHON_EXTRA_COMPILE_ARGS "${GUDHI_PYTHON_EXTRA_COMPILE_ARGS}'-DBOOST_RESULT_OF_USE_DECLTYPE', ") set(GUDHI_PYTHON_EXTRA_COMPILE_ARGS "${GUDHI_PYTHON_EXTRA_COMPILE_ARGS}'-DBOOST_ALL_NO_LIB', ") @@ -496,11 +499,17 @@ if(PYTHONINTERP_FOUND) # Wasserstein if(OT_FOUND AND PYBIND11_FOUND) - if(TORCH_FOUND AND EAGERPY_FOUND) + # EagerPy dependency because of enable_autodiff=True + if(EAGERPY_FOUND) add_gudhi_py_test(test_wasserstein_distance) endif() add_gudhi_py_test(test_wasserstein_barycenter) endif() + if(OT_FOUND) + if(TORCH_FOUND AND TENSORFLOW_FOUND AND EAGERPY_FOUND) + add_gudhi_py_test(test_wasserstein_with_tensors) + endif() + endif() # Representations if(SKLEARN_FOUND AND MATPLOTLIB_FOUND) diff --git a/src/python/doc/installation.rst b/src/python/doc/installation.rst index 78e1af73..66efe45a 100644 --- a/src/python/doc/installation.rst +++ b/src/python/doc/installation.rst @@ -40,7 +40,7 @@ different, and in particular the `python/` subdirectory is actually `src/python/ there. The library uses c++14 and requires `Boost <https://www.boost.org/>`_ :math:`\geq` 1.56.0, -`CMake <https://www.cmake.org/>`_ :math:`\geq` 3.1 to generate makefiles, +`CMake <https://www.cmake.org/>`_ :math:`\geq` 3.5 to generate makefiles, `NumPy <http://numpy.org>`_, `Cython <https://www.cython.org/>`_ and `pybind11 <https://github.com/pybind/pybind11>`_ to compile the GUDHI Python module. @@ -65,7 +65,7 @@ one can build the GUDHI Python module, by running the following commands in a te cd /path-to-gudhi/ mkdir build cd build/ - cmake .. + cmake -DCMAKE_BUILD_TYPE=Release .. cd python make @@ -394,6 +394,11 @@ mathematics, science, and engineering. :class:`~gudhi.point_cloud.knn.KNearestNeighbors` can use the Python package `SciPy <http://scipy.org>`_ as a backend if explicitly requested. +TensorFlow +---------- + +`TensorFlow <https://www.tensorflow.org>`_ is currently only used in some automatic differentiation tests. + Bug reports and contributions ***************************** diff --git a/src/python/doc/rips_complex_sum.inc b/src/python/doc/rips_complex_sum.inc index c123ea2a..2cb24990 100644 --- a/src/python/doc/rips_complex_sum.inc +++ b/src/python/doc/rips_complex_sum.inc @@ -1,14 +1,14 @@ .. table:: :widths: 30 40 30 - +----------------------------------------------------------------+------------------------------------------------------------------------+----------------------------------------------------------------------+ - | .. figure:: | The Vietoris-Rips complex is a simplicial complex built as the | :Authors: Clément Maria, Pawel Dlotko, Vincent Rouvreau, Marc Glisse | - | ../../doc/Rips_complex/rips_complex_representation.png | clique-complex of a proximity graph. | | - | :figclass: align-center | | :Since: GUDHI 2.0.0 | - | | We also provide sparse approximations, to speed-up the computation | | - | | of persistent homology, and weighted versions, which are more robust | :License: MIT | - | | to outliers. | | - | | | | - +----------------------------------------------------------------+------------------------------------------------------------------------+----------------------------------------------------------------------+ - | * :doc:`rips_complex_user` | * :doc:`rips_complex_ref` | - +----------------------------------------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------+ + +----------------------------------------------------------------+------------------------------------------------------------------------+----------------------------------------------------------------------------------+ + | .. figure:: | The Vietoris-Rips complex is a simplicial complex built as the | :Authors: Clément Maria, Pawel Dlotko, Vincent Rouvreau, Marc Glisse, Yuichi Ike | + | ../../doc/Rips_complex/rips_complex_representation.png | clique-complex of a proximity graph. | | + | :figclass: align-center | | :Since: GUDHI 2.0.0 | + | | We also provide sparse approximations, to speed-up the computation | | + | | of persistent homology, and weighted versions, which are more robust | :License: MIT | + | | to outliers. | | + | | | | + +----------------------------------------------------------------+------------------------------------------------------------------------+----------------------------------------------------------------------------------+ + | * :doc:`rips_complex_user` | * :doc:`rips_complex_ref` | + +----------------------------------------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------+ diff --git a/src/python/doc/rips_complex_user.rst b/src/python/doc/rips_complex_user.rst index 6048cc4e..27d218d4 100644 --- a/src/python/doc/rips_complex_user.rst +++ b/src/python/doc/rips_complex_user.rst @@ -7,9 +7,9 @@ Rips complex user manual Definition ---------- -==================================================================== ================================ ====================== -:Authors: Clément Maria, Pawel Dlotko, Vincent Rouvreau, Marc Glisse :Since: GUDHI 2.0.0 :License: GPL v3 -==================================================================== ================================ ====================== +================================================================================ ================================ ====================== +:Authors: Clément Maria, Pawel Dlotko, Vincent Rouvreau, Marc Glisse, Yuichi Ike :Since: GUDHI 2.0.0 :License: GPL v3 +================================================================================ ================================ ====================== +-------------------------------------------+----------------------------------------------------------------------+ | :doc:`rips_complex_user` | :doc:`rips_complex_ref` | diff --git a/src/python/gudhi/simplex_tree.pxd b/src/python/gudhi/simplex_tree.pxd index 6ec85bca..3c4cbed3 100644 --- a/src/python/gudhi/simplex_tree.pxd +++ b/src/python/gudhi/simplex_tree.pxd @@ -64,6 +64,7 @@ cdef extern from "Simplex_tree_interface.h" namespace "Gudhi": void compute_extended_filtration() nogil vector[vector[pair[int, pair[double, double]]]] compute_extended_persistence_subdiagrams(vector[pair[int, pair[double, double]]] dgm, double min_persistence) nogil Simplex_tree_interface_full_featured* collapse_edges(int nb_collapse_iteration) nogil + void reset_filtration(double filtration, int dimension) nogil # Iterators over Simplex tree pair[vector[int], double] get_simplex_and_filtration(Simplex_tree_simplex_handle f_simplex) nogil Simplex_tree_simplices_iterator get_simplices_iterator_begin() nogil diff --git a/src/python/gudhi/simplex_tree.pyx b/src/python/gudhi/simplex_tree.pyx index 5043c621..c671da56 100644 --- a/src/python/gudhi/simplex_tree.pyx +++ b/src/python/gudhi/simplex_tree.pyx @@ -342,7 +342,7 @@ cdef class SimplexTree: return self.get_ptr().prune_above_filtration(filtration) def expansion(self, max_dim): - """Expands the Simplex_tree containing only its one skeleton + """Expands the simplex tree containing only its one skeleton until dimension max_dim. The expanded simplicial complex until dimension :math:`d` @@ -352,7 +352,7 @@ cdef class SimplexTree: The filtration value assigned to a simplex is the maximal filtration value of one of its edges. - The Simplex_tree must contain no simplex of dimension bigger than + The simplex tree must contain no simplex of dimension bigger than 1 when calling the method. :param max_dim: The maximal dimension. @@ -372,6 +372,20 @@ cdef class SimplexTree: """ return self.get_ptr().make_filtration_non_decreasing() + def reset_filtration(self, filtration, min_dim = 0): + """This function resets the filtration value of all the simplices of dimension at least min_dim. Resets all the + simplex tree when `min_dim = 0`. + `reset_filtration` may break the filtration property with `min_dim > 0`, and it is the user's responsibility to + make it a valid filtration (using a large enough `filt_value`, or calling `make_filtration_non_decreasing` + afterwards for instance). + + :param filtration: New threshold value. + :type filtration: float. + :param min_dim: The minimal dimension. Default value is 0. + :type min_dim: int. + """ + self.get_ptr().reset_filtration(filtration, min_dim) + def extend_filtration(self): """ Extend filtration for computing extended persistence. This function only uses the filtration values at the 0-dimensional simplices, and computes the extended persistence @@ -380,14 +394,14 @@ cdef class SimplexTree: .. note:: Note that after calling this function, the filtration - values are actually modified within the Simplex_tree. + values are actually modified within the simplex tree. The function :func:`extended_persistence` retrieves the original values. .. note:: Note that this code creates an extra vertex internally, so you should make sure that - the Simplex_tree does not contain a vertex with the largest possible value (i.e., 4294967295). + the simplex tree does not contain a vertex with the largest possible value (i.e., 4294967295). """ self.get_ptr().compute_extended_filtration() diff --git a/src/python/gudhi/subsampling.pyx b/src/python/gudhi/subsampling.pyx index f77c6f75..b11d07e5 100644 --- a/src/python/gudhi/subsampling.pyx +++ b/src/python/gudhi/subsampling.pyx @@ -33,7 +33,7 @@ def choose_n_farthest_points(points=None, off_file='', nb_points=0, starting_poi The iteration starts with the landmark `starting point`. :param points: The input point set. - :type points: Iterable[Iterable[float]]. + :type points: Iterable[Iterable[float]] Or @@ -42,14 +42,15 @@ def choose_n_farthest_points(points=None, off_file='', nb_points=0, starting_poi And in both cases - :param nb_points: Number of points of the subsample. - :type nb_points: unsigned. + :param nb_points: Number of points of the subsample (the subsample may be \ + smaller if there are fewer than nb_points distinct input points) + :type nb_points: int :param starting_point: The iteration starts with the landmark `starting \ - point`,which is the index of the point to start with. If not set, this \ + point`, which is the index of the point to start with. If not set, this \ index is chosen randomly. - :type starting_point: unsigned. + :type starting_point: int :returns: The subsample point set. - :rtype: List[List[float]]. + :rtype: List[List[float]] """ if off_file: if os.path.isfile(off_file): @@ -76,7 +77,7 @@ def pick_n_random_points(points=None, off_file='', nb_points=0): """Subsample a point set by picking random vertices. :param points: The input point set. - :type points: Iterable[Iterable[float]]. + :type points: Iterable[Iterable[float]] Or @@ -86,7 +87,7 @@ def pick_n_random_points(points=None, off_file='', nb_points=0): And in both cases :param nb_points: Number of points of the subsample. - :type nb_points: unsigned. + :type nb_points: int :returns: The subsample point set. :rtype: List[List[float]] """ @@ -107,7 +108,7 @@ def sparsify_point_set(points=None, off_file='', min_squared_dist=0.0): between any two points is greater than or equal to min_squared_dist. :param points: The input point set. - :type points: Iterable[Iterable[float]]. + :type points: Iterable[Iterable[float]] Or @@ -118,7 +119,7 @@ def sparsify_point_set(points=None, off_file='', min_squared_dist=0.0): :param min_squared_dist: Minimum squared distance separating the output \ points. - :type min_squared_dist: float. + :type min_squared_dist: float :returns: The subsample point set. :rtype: List[List[float]] """ diff --git a/src/python/test/test_simplex_tree.py b/src/python/test/test_simplex_tree.py index d79d2c34..3b23fa0b 100755 --- a/src/python/test/test_simplex_tree.py +++ b/src/python/test/test_simplex_tree.py @@ -359,6 +359,28 @@ def test_collapse_edges(): for simplex in st.get_skeleton(0): assert simplex[1] == 1. +def test_reset_filtration(): + st = SimplexTree() + + assert st.insert([0, 1, 2], 3.) == True + assert st.insert([0, 3], 2.) == True + assert st.insert([3, 4, 5], 3.) == True + assert st.insert([0, 1, 6, 7], 4.) == True + + # Guaranteed by construction + for simplex in st.get_simplices(): + assert st.filtration(simplex[0]) >= 2. + + # dimension until 5 even if simplex tree is of dimension 3 to test the limits + for dimension in range(5, -1, -1): + st.reset_filtration(0., dimension) + for simplex in st.get_skeleton(3): + print(simplex) + if len(simplex[0]) < (dimension) + 1: + assert st.filtration(simplex[0]) >= 2. + else: + assert st.filtration(simplex[0]) == 0. + def test_boundaries_iterator(): st = SimplexTree() diff --git a/src/python/test/test_wasserstein_distance.py b/src/python/test/test_wasserstein_distance.py index 90d26809..e3b521d6 100755 --- a/src/python/test/test_wasserstein_distance.py +++ b/src/python/test/test_wasserstein_distance.py @@ -97,27 +97,3 @@ def test_wasserstein_distance_pot(): def test_wasserstein_distance_hera(): _basic_wasserstein(hera_wrap(delta=1e-12), 1e-12, test_matching=False) _basic_wasserstein(hera_wrap(delta=.1), .1, test_matching=False) - -def test_wasserstein_distance_grad(): - import torch - - diag1 = torch.tensor([[2.7, 3.7], [9.6, 14.0], [34.2, 34.974]], requires_grad=True) - diag2 = torch.tensor([[2.8, 4.45], [9.5, 14.1]], requires_grad=True) - diag3 = torch.tensor([[2.8, 4.45], [9.5, 14.1]], requires_grad=True) - assert diag1.grad is None and diag2.grad is None and diag3.grad is None - dist12 = pot(diag1, diag2, internal_p=2, order=2, enable_autodiff=True) - dist30 = pot(diag3, torch.tensor([]), internal_p=2, order=2, enable_autodiff=True) - dist12.backward() - dist30.backward() - assert not torch.isnan(diag1.grad).any() and not torch.isnan(diag2.grad).any() and not torch.isnan(diag3.grad).any() - diag4 = torch.tensor([[0., 10.]], requires_grad=True) - diag5 = torch.tensor([[1., 11.], [3., 4.]], requires_grad=True) - dist45 = pot(diag4, diag5, internal_p=1, order=1, enable_autodiff=True) - assert dist45 == 3. - dist45.backward() - assert np.array_equal(diag4.grad, [[-1., -1.]]) - assert np.array_equal(diag5.grad, [[1., 1.], [-1., 1.]]) - diag6 = torch.tensor([[5., 10.]], requires_grad=True) - pot(diag6, diag6, internal_p=2, order=2, enable_autodiff=True).backward() - # https://github.com/jonasrauber/eagerpy/issues/6 - # assert np.array_equal(diag6.grad, [[0., 0.]]) diff --git a/src/python/test/test_wasserstein_with_tensors.py b/src/python/test/test_wasserstein_with_tensors.py new file mode 100755 index 00000000..e3f1411a --- /dev/null +++ b/src/python/test/test_wasserstein_with_tensors.py @@ -0,0 +1,47 @@ +""" This file is part of the Gudhi Library - https://gudhi.inria.fr/ - which is released under MIT. + See file LICENSE or go to https://gudhi.inria.fr/licensing/ for full license details. + Author(s): Mathieu Carriere + + Copyright (C) 2020 Inria + + Modification(s): + - YYYY/MM Author: Description of the modification +""" + +from gudhi.wasserstein import wasserstein_distance as pot +import numpy as np +import torch +import tensorflow as tf + +def test_wasserstein_distance_grad(): + diag1 = torch.tensor([[2.7, 3.7], [9.6, 14.0], [34.2, 34.974]], requires_grad=True) + diag2 = torch.tensor([[2.8, 4.45], [9.5, 14.1]], requires_grad=True) + diag3 = torch.tensor([[2.8, 4.45], [9.5, 14.1]], requires_grad=True) + assert diag1.grad is None and diag2.grad is None and diag3.grad is None + dist12 = pot(diag1, diag2, internal_p=2, order=2, enable_autodiff=True) + dist30 = pot(diag3, torch.tensor([]), internal_p=2, order=2, enable_autodiff=True) + dist12.backward() + dist30.backward() + assert not torch.isnan(diag1.grad).any() and not torch.isnan(diag2.grad).any() and not torch.isnan(diag3.grad).any() + diag4 = torch.tensor([[0., 10.]], requires_grad=True) + diag5 = torch.tensor([[1., 11.], [3., 4.]], requires_grad=True) + dist45 = pot(diag4, diag5, internal_p=1, order=1, enable_autodiff=True) + assert dist45 == 3. + dist45.backward() + assert np.array_equal(diag4.grad, [[-1., -1.]]) + assert np.array_equal(diag5.grad, [[1., 1.], [-1., 1.]]) + diag6 = torch.tensor([[5., 10.]], requires_grad=True) + pot(diag6, diag6, internal_p=2, order=2, enable_autodiff=True).backward() + # https://github.com/jonasrauber/eagerpy/issues/6 + # assert np.array_equal(diag6.grad, [[0., 0.]]) + +def test_wasserstein_distance_grad_tensorflow(): + with tf.GradientTape() as tape: + diag4 = tf.convert_to_tensor(tf.Variable(initial_value=np.array([[0., 10.]]), trainable=True)) + diag5 = tf.convert_to_tensor(tf.Variable(initial_value=np.array([[1., 11.], [3., 4.]]), trainable=True)) + dist45 = pot(diag4, diag5, internal_p=1, order=1, enable_autodiff=True) + assert dist45 == 3. + + grads = tape.gradient(dist45, [diag4, diag5]) + assert np.array_equal(grads[0].values, [[-1., -1.]]) + assert np.array_equal(grads[1].values, [[1., 1.], [-1., 1.]])
\ No newline at end of file |