FastPareto

Fast implementations for finding the pareto front in a large set of points

Done:

• A very quick python-numpy implementation for finding pareto points out of a set of 2d points:

  import numpy as np
  from fastpareto import pareto
  import timeit
  testdata = np.random.randn(1000000,2) #one million test points
  t = timeit.timeit('pareto(testdata)', globals=globals(),number=10)/10
  print(str(t)+' seconds for finding pareto points in a set of 1 million 2d points')
  #>> 0.2959234861191362 seconds for finding pareto points in a set of 1 million 2d points

• A less quick python-numpy implemention for more than 2 costs:

  import numpy as np
  from fastpareto import pareto
  import timeit
  testdata = np.random.randn(1000000,3) #one million test points
  t = timeit.timeit('pareto(testdata)', globals=globals(),number=10)/10
  print(str(t)+' seconds for finding pareto points in a set of 1 million 3d points')
  #>> 6.359629309689626 seconds for finding pareto points in a set of 1 million 3d points

  In this case, Peter's answer on StackOverflow seems to be quicker and is thus recommended.

Todo:

• Faster python implementation for more than 2 costs
• Matlab port

Legal:

• License (including disclaimers): see the LICENSE file, which is a copy of the GNU GPL license.