const training = [
  [0,3], [1,6], [2,8], [3,7], [9, 100]
]

// alpha
const learningRate = 0.05

const hypothesis = (x,params) => params[1] * x + params[0]

const cost = (training,params) => {
  let sum = 0;

  for (var i = 0; i < training.length; i++) {
    sum += Math.pow(hypothesis(training[i][0],params) - training[i][1], 2);
  }
  console.log(sum/(2*training.length))
  return sum / (2 * training.length);
}
 // summed error needs to be generalized
const summedError1 = (training,params) => {
  let sum1 = 0
  for(let i=0; i<training.length; i++){
    sum1 += hypothesis(training[i][0],params) - training[i][1]
  }
  return sum1 / training.length
}
const summedError2 = (training,params) => {
  let sum2 = 0
  for(let i=0; i<training.length; i++){
    sum2 += (hypothesis(training[i][0], params) - training[i][1]) * training[i][0]
  }
  return sum2/training.length
}
//  generalized
const summedError = (training,params) => {
  let sum = 0
  for(let i=0; i<training.length; i++){

  }
  return sum / training.length
}

// this function needs refactoring allow for multiple temp and param variables [i][j]
const gradientDescent = (training, params, learningRate, iterations) => {
  let count = 0
  while (count < iterations) {
    let temp0 = params[0] - (learningRate * summedError1(training,params))
    let temp1 = params[1] - (learningRate * summedError2(training,params))
    params[0] = temp0
    params[1] = temp1
    count++
    cost(training,params)
  }
  console.log(params)
}
gradientDescent(training,[0,0],learningRate,1000)