const math = require("mathjs");
/**
 *  @typedef {[number]} inputVector
 *
 */
/**
 * @typedef {number} expectedOutput
 */

/**
 * @param {number} learningRate - between 0 and 1 (larger values make the weight change more volatile)
 * @param trainingSet {[{inputVector,expectedOutput,song}]}
 */
function perceptron(trainingSet) {
  const learningRate = 1;
  console.log(trainingSet);
  const inputsWithBias = trainingSet.map(item => [1, ...item.inputVector]);
  const weights = new Array(inputsWithBias[0].length).fill(0);
  const trainedNetwork = inputsWithBias.reduce(
    (updatedWeights, curInput, ind) => {
      const output = math.multiply(updatedWeights, curInput);
      const newWeights = weights.map(
        (weight, i) =>
          weight +
          learningRate *
            (trainingSet[ind].expectedOutput - output) *
            curInput[i]
      );
      return newWeights;
    },
    weights
  );
  console.log(trainedNetwork);
  return trainedNetwork;
  // console.log(wegihts)
  // math.multiply(weights,trainingSet)
}

const trainedNetwork = perceptron([
  { inputVector: [6], expectedOutput: 1 },
  { inputVector: [-5], expectedOutput: -1 },
  { inputVector: [10], expectedOutput: 1 },
  { inputVector: [-2], expectedOutput: 1 }
]);