stream2.tl.extend_leaves

stream2.tl.extend_leaves(adata, Mode='QuantDists', ControlPar=0.5, DoSA=True, DoSA_maxiter=200, LeafIDs=None, TrimmingRadius=inf, key='epg')[source]

Extend leaves with additional nodes

Parameters:

  • Mode (str, the mode used to extend the graph.) – “QuantDists”,”QuantCentroid”, “WeigthedCentroid”

  • LeafIDs (int vector,) – The id of nodes to extend. If None, all the vertices will be extended.

  • TrimmingRadius (positive numeric) – The trimming radius used to control distance

  • DoSA (bool) – Should optimization (via simulated annealing) be performed when Mode = “QuantDists”?

  • ControlPar (positive numeric) –

    The parameter used to control the contribution of the different data points

    The value of ControlPar has a different interpretation depending on the valus of Mode. In each case, for only the extreme points, i.e., the points associated with the leaf node that do not have a projection on any edge are considered.

    If Mode = “QuantCentroid”, for each leaf node, the extreme points are ordered by their distance from the node and the centroid of the points farther away than ControlPar is returned.

    If Mode = “WeightedCentroid”, for each leaf node, a weight is computed for each points by raising the distance to the ControlPar power. Hence, larger values of ControlPar result in a larger influence of points farther from the node