Geo

API details

DiffusionDistance

 DiffusionDistance (t_max=5, knn=5, anisotropy=1, log=False,
                    normalize=False, symmetrize=False)

class DiffusionDistance
X (np.array) data t_max (int), 2^t_max is the max scale of the Diffusion kernel knn (int) = 5 number of neighbors for the KNN in the alpha decay kernel construction, same default as in PHATE Anisotropy (int): the alpha in Coifman Lafon 2006, 1: double normalization 0: usual random walk log (bool) log(P) or not normalize (bool) min-max normalization of the distance matrix phate (bool) is PHATE op if true (should be the same as graphtool)


DiffusionAffinity

 DiffusionAffinity (knn=5, anisotropy=0, t_diff=1, topeig=100)

class DiffusionAffinity
X (np.array) data t_max (int), 2^t_max is the max scale of the Diffusion kernel knn (int) = 5 number of neighbors for the KNN in the alpha decay kernel construction, same default as in PHATE Anisotropy (int): the alpha in Coifman Lafon 2006, 1: double normalization 0: usual random walk t_diff (int) the power of the diffusion affinity matrix topeig (int) in the the top k eigenvalues to consider in the spectral decomposition

return the l2 between the row of the affinity matrix A^t


DiffusionMap

 DiffusionMap (knn=5, anisotropy=0, t_diff=1, topeig=100, n_emb=10)

class DiffusionMap
X (np.array) data t_max (int), 2^t_max is the max scale of the Diffusion kernel knn (int) = 5 number of neighbors for the KNN in the alpha decay kernel construction, same default as in PHATE Anisotropy (int): the alpha in Coifman Lafon 2006, 1: double normalization 0: usual random walk t_diff (int) the power of the diffusion affinity matrix topeig (int) in the the top k eigenvalues to consider in the spectral decomposition n_emb (int) the dimension of the emb space

return the pairwise dist. in the diffusion map embedding space


PhateDistance

 PhateDistance (knn=5, anisotropy=0, verbose=False)

class PhateDistance
X (np.array) data knn (int) = 5 number of neighbors for the KNN in the alpha decay kernel construction, same default as in PHATE Anisotropy (int): the alpha in Coifman Lafon 2006, 1: double normalization 0: usual random walk verbose (bool): verbose param. in PHATE

return PHATE distance the L2 between Potential of Heat-diffusion


old_DiffusionDistance

 old_DiffusionDistance (kernel, t_max)

Initialize self. See help(type(self)) for accurate signature.


setup_distance

 setup_distance (distance_type:str='gaussian', rbf_length_scale:float=0.5,
                 t_max:int=5, knn:int=5)