Nelder-Mead simplex method implementation on js

Forked from SimpleSimplex

Added 2D optimization support

Rewritten in ES6 module format as js class

Added objective function memoization, that reduce objFunc calls by ~6 times

Added mocha tests

Example

Create solver object

import Simplex from '../src/NMSimplex'
const s = new Simplex()

Create objective function, that take an array of 2 numbers for 2D optimization or number for 1D optimization and return number

function objFunc (arr) {
  return (5 * arr[0] + 1) ** 2 + (4 * arr[1] - 16) ** 2
}

Optionally you can create callback function that solver will be call in the end of each iteration. For example it can be function that push each iteration in array

const rez = []
function callback (
  itr, // number of iteration, umber
  x, // best vertex, number or array of 2 numbers
  obj, // objFunc(x), number
  centr, // centroid coordinates, number or array of 2 numbers
  action, // iteration action, string
  objEvals // number of objFunc calls, number
) {
  rez.push([itr, x, obj, centr, action, objEvals])
}

const result = s.solve(
  objFunc,
  [-354, 1153], // x0 - initial vertex, number or array of 2 numbers
  1000, // maximum number of iterations, number
  callback, // callback function (optional)
  true // use memoization (optional, default true), boolean
)
// [[-0.2, 4], best vertex
// 0, objFunc(x)
// [-0.2, 4], centroid coordinates
// "tolerance reached"] last action or 'tolerance reached'