kd tree build sort optimize

Question

The algorithm build of kd-tree implemented in the Python programming language is as follows (from http://en.wikipedia.org/wiki/K-d_tree):

class Node: pass

def kdtree(point_list, depth=0):

  if not point_list:
    return None

  # Select axis based on depth so that axis cycles through all valid values
  k = len(point_list[0]) # assumes all points have the same dimension
  axis = depth % k

  # Sort point list and choose median as pivot element
  point_list.sort(key=lambda point: point[axis])
  median = len(point_list) // 2 # choose median

  # Create node and construct subtrees
  node = Node()
  node.location = point_list[median]
  node.left_child = kdtree(point_list[:median], depth + 1)
  node.right_child = kdtree(point_list[median + 1:], depth + 1)
  return node

Sorting is performed on every step. How to reduce the amount of sorting?

you must sorting array on every step, complexity quick sort nlog(n) => complexity tree build k*nlog(n), where k - number of steps; i want optimize algorithm tree build — Fandorin, Sep 19 '12 at 13:58

score 1 · Accepted Answer · answered Sep 19 '12 at 12:25

1

It looks like you're only sorting to split around the median. Instead, you could implement a linear-time selection algorithm such as quickselect, then do a linear-time partition of point_list. Then, you don't need to sort at all anymore.

answered Sep 19 '12 at 12:25

Fred Foo

355,277
75
744
836

Thanks! I googled http://ndevilla.free.fr/median/median/index.html (quickselect.c). I can change sort on quickselect method. Yes? – Fandorin Sep 19 '12 at 13:14
@Fandorin: yes, if you then do the splitting in a different way. `point_list[:median]` depends on the sorting order. – Fred Foo Sep 19 '12 at 14:23

kd tree build sort optimize

1 Answers1