Regarding my comment in PR #43 (see https://github.com/epfl-lts2/pygsp/pull/43#discussion_r336029429), I was thinking about something like that:
when querying with a radius, use tree.query_ball_point instead of query such that you don't allocate a distances matrix of size NxN.
I ran the test_nearest_neighbor locally and it seems it is passing.
Regarding my comment in PR #43 (see https://github.com/epfl-lts2/pygsp/pull/43#discussion_r336029429), I was thinking about something like that: when querying with a radius, use
tree.query_ball_pointinstead ofquerysuch that you don't allocate a distances matrix of size NxN.I ran the
test_nearest_neighborlocally and it seems it is passing.