I would like to randomly select one element from an array, but each element has a known probability of selection.
All chances together (within the array) sums to 1.<
Another Ruby example:
def weighted_rand(weights = {})
raise 'Probabilities must sum up to 1' unless weights.values.inject(&:+) == 1.0
raise 'Probabilities must not be negative' unless weights.values.all? { |p| p >= 0 }
# Do more sanity checks depending on the amount of trust in the software component using this method
# E.g. don't allow duplicates, don't allow non-numeric values, etc.
# Ignore elements with probability 0
weights = weights.reject { |k, v| v == 0.0 } # e.g. => {"a"=>0.4, "b"=>0.4, "c"=>0.2}
# Accumulate probabilities and map them to a value
u = 0.0
ranges = weights.map { |v, p| [u += p, v] } # e.g. => [[0.4, "a"], [0.8, "b"], [1.0, "c"]]
# Generate a (pseudo-)random floating point number between 0.0(included) and 1.0(excluded)
u = rand # e.g. => 0.4651073966724186
# Find the first value that has an accumulated probability greater than the random number u
ranges.find { |p, v| p > u }.last # e.g. => "b"
end
How to use:
weights = {'a' => 0.4, 'b' => 0.4, 'c' => 0.2, 'd' => 0.0}
weighted_rand weights
What to expect roughly:
sample = 1000.times.map{ weighted_rand weights }
sample.count('a') # 396
sample.count('b') # 406
sample.count('c') # 198
sample.count('d') # 0