I have a homework to do, that is:
Pick a random color, from yellow, blue and red, given the probability of:
Yellow: 3/7
Blue: 1/7
Red: 3/7
I know that I could work this around by using something like:
[yellow, yellow, yellow, blue, red, red, red]
But I don’t think this would be programatically good, since when I chance the probability, I would have to change the array.
So, I thought I could try something like a weight approach
let yellow_probability = 3/7
let blue_probability = 1/7
let red_probability = 3/7
const colors = ['yellow', 'blue', 'red']
function pickPosition(yellow_probability, blue_probability, red_probability){
let yellow_weight = Math.random() * yellow_probability
let blue_weight = Math.random() * blue_probability
let red_weight = Math.random() * red_probability
let weights = [yellow_weight, blue_weight, red_weight]
let max_of_array = Math.max.apply(Math, weights);
pickedColor = weights.indexOf(max_of_array)
return pickedColor
}
pickedColorIndex = pickPosition(yellow_probability, blue_probability, red_probability)
pickedColor = colors[pickedColorIndex]
console.log(pickedColor)
I did a test:
let n=1000000;
let yellow=0, blue=0, red=0;
for (let i=0; i<n; i++) {
pickedColorIndex = pickPosition(yellow_probability, blue_probability, red_probability)
if (pickedColorIndex==0) yellow++
else if (pickedColorIndex==1) blue++
else red++;
}
console.log("yellow = " + yellow/n );
console.log("blue = " + blue/n );
console.log("red = " + red/n );
And I would expect this test to output something like:
Yellow = 0.43
Blue = 0.14
Red = 0.43
But I am getting:
Yellow = 0.48
Blue = 0.03
Red = 0.48
It is interesting to point out that the code works when the probabilities are equal (1/3, 1/3, 1/3) or something like (1/2, 1/2, 0)
Can anyone point out what I am doing wrong?
This is similar to the approach mentioned in the duplicate. You create an array with ratios same as the probabilities. (Here I’m using 2 decimal places and adding ~100 items to the array. You could add multiply by a bigger number and have .toFixed(3) for better accuracy)
function getRandomWithProbability(array) {
const filled = array.flatMap(([color, prob]) => {
const length = prob.toFixed(2) * 100;
return Array.from({ length }).fill(color)
});
const random = Math.floor(Math.random() * filled.length);
return filled[random]
}
const arr = [["yellow", 3/7], ["blue", 1/7], ["red", 3/7]]
console.log(getRandomWithProbability(arr))
Instead of a single random value, you create as many as different items you have and later take the one with the max value.
This promotes values/items with a higher factor/probability.
Instead of this approach, you could take a single random value and take all probabilities into an array and check in which interval the random value is. Take this item.
EDIT: The code
function getRandomIndex(probabilities) {
var random = Math.random(),
i;
for (i = 0; i < probabilities.length; i++) {
if (random < probabilities[i]) return i;
random -= probabilities[i];
}
return probabilites.length - 1;
}
var probabilities = [3 / 7, 1 / 7, 3 / 7],
j = 1e6,
count = [0, 0, 0];
while (j--) count[getRandomIndex(probabilities)]++;
console.log(count);