Entropy-Calculation

entropy.js

/*
Calculates entropy and conditional entropy as defined in Information Theory (https://en.wikipedia.org/wiki/Entropy_(information_theory))

entropy = -Sum(P_i*log_2(P_i))
*/

const Entropy = (...values) => {
  let calculationResult = 0;
  let fraction = 0;

  if (!values) {
    throw new Error('no parameters entered!');
  }
  const valuesSum = values.reduce((sum, curr) => sum + curr, 0);

  for (let v of values) {
    if (v !== 0) {
      fraction = v/valuesSum;
      calculationResult += fraction * Math.log2(fraction);
    }
  }

  return -1 * calculationResult;
}

/*
Conditional Entropy

value is of format:
{
  fraction: 2/3,
  entropyTuple: [1, 2, 3]
}
 */
const Info = (...values) => {
  let calculationResult = 0;

  if (!values) {
    throw new Error('no parameters entered!');
  }

  for (let v of values) {
    let { fraction, entropyTuple } = v;
    if (!entropyTuple) {
      throw new Error('illegal entropyTuple!');
    }

    calculationResult += fraction * Entropy(...entropyTuple);
  }

  return calculationResult;
}

// example:
const entropy = Entropy(8,2,4); // 1.37
const info = Info({ fraction: 6/14, entropyTuple: [1,0,5]},
  { fraction: 4/14, entropyTuple: [1,2,1]},
  { fraction: 4/14, entropyTuple: [2,0,2]}); // 0.992