Netlogo: Assign variable using probabilities

问题:

How to assign a string or integer variable to turtle, using probabilities of the variables in a group/list? For example it is 0.4 probability that one specific variable is used from specific group/list. The function selects randomly the variable based on probability. I need to use the same method afterwards to choose a variable (string) from a list according to probability. In python it should be:

import random
def random_value(probability_list, values):
    r = random.random()
    index = 0
    while(r >= 0 and index < len(probability_list)):
      r -= probability_list[index]
      index += 1
    value=values[index - 1]
    value_index=index-1
    return value,value_index

I tried it in Netlogo like below (get error that index is -1) but is there a better way?

globals [random_nr probabilities some_list index]
to initialize-variables
  set some_list[]
  set probabilities[]
end
to random_pick
  set random_nr random-float 1
  set probabilities [0.1 0.2 0.4 0.3]
  set some_list ["String1" "String2" "String3" "String4"]
  set index 0
  while [(random_nr >= 0) and (length probabilities < index)] [
   set random_nr random_nr - item index probabilities
   set index index + 1 ]
  set index index - 1
end

回答1:

is there a better way?

Yes there is.

NetLogo 6.0 comes with the rnd extension bundled. (You can also download the extension separately for earlier versions of NetLogo.)

The rnd extension offers the rnd:weighted-one-of-list primitive, which does exactly what you're trying to do:

extensions [ rnd ]

to-report pick
  let probabilities [0.1 0.2 0.4 0.3]
  let some_list ["String1" "String2" "String3" "String4"]
  report first rnd:weighted-one-of-list (map list some_list probabilities) last
end

Let me unpack the last expression a bit:

• The role of (map list some_list probabilities) is to "zip" the two lists together, in order to get a list of pairs of the form: [["String1" 0.1] ["String2" 0.2] ["String3" 0.4] ["String4" 0.3]].

• That list of pairs is passed as the first argument to rnd:weighted-one-of-list. We pass last as the second argument of rnd:weighted-one-of-list to tell it that it should use the second item of each pair as the probability.

• rnd:weighted-one-of-list then picks one of the pairs at random, and returns that whole pair. But since we're only interested in the first item of the pair, we use first to extract it.

To understand how that code works, it helps to understand how Anonymous procedures work. Note how we make use of the concise syntax for passing list to map and for passing last to rnd:weighted-one-of-list.

回答2:

Don't overlook the rnd extension: https://github.com/NetLogo/Rnd-Extension But it is possible to do it essentially as you propose. I'll to that here, but it would be better to use explicit arguments.

to-report random-pick
  let _r random-float 1
  let _ps [0.1 0.2 0.4 0.3]
  let _lst ["String1" "String2" "String3" "String4"]
  let _i 0
  while [_r >= item _i _ps] [
   set _r (_r - item _i _ps)
   set _i (_i + 1) ]
  report item _i _lst
end