simple-squiggle

A restricted subset of Squiggle
Log | Files | Refs | README

erf.js (5471B)


      1 "use strict";
      2 
      3 Object.defineProperty(exports, "__esModule", {
      4   value: true
      5 });
      6 exports.createErf = void 0;
      7 
      8 var _collection = require("../../utils/collection.js");
      9 
     10 var _number = require("../../utils/number.js");
     11 
     12 var _factory = require("../../utils/factory.js");
     13 
     14 /* eslint-disable no-loss-of-precision */
     15 var name = 'erf';
     16 var dependencies = ['typed'];
     17 var createErf = /* #__PURE__ */(0, _factory.factory)(name, dependencies, function (_ref) {
     18   var typed = _ref.typed;
     19 
     20   /**
     21    * Compute the erf function of a value using a rational Chebyshev
     22    * approximations for different intervals of x.
     23    *
     24    * This is a translation of W. J. Cody's Fortran implementation from 1987
     25    * ( https://www.netlib.org/specfun/erf ). See the AMS publication
     26    * "Rational Chebyshev Approximations for the Error Function" by W. J. Cody
     27    * for an explanation of this process.
     28    *
     29    * For matrices, the function is evaluated element wise.
     30    *
     31    * Syntax:
     32    *
     33    *    math.erf(x)
     34    *
     35    * Examples:
     36    *
     37    *    math.erf(0.2)    // returns 0.22270258921047847
     38    *    math.erf(-0.5)   // returns -0.5204998778130465
     39    *    math.erf(4)      // returns 0.9999999845827421
     40    *
     41    * @param {number | Array | Matrix} x   A real number
     42    * @return {number | Array | Matrix}    The erf of `x`
     43    */
     44   return typed('name', {
     45     number: function number(x) {
     46       var y = Math.abs(x);
     47 
     48       if (y >= MAX_NUM) {
     49         return (0, _number.sign)(x);
     50       }
     51 
     52       if (y <= THRESH) {
     53         return (0, _number.sign)(x) * erf1(y);
     54       }
     55 
     56       if (y <= 4.0) {
     57         return (0, _number.sign)(x) * (1 - erfc2(y));
     58       }
     59 
     60       return (0, _number.sign)(x) * (1 - erfc3(y));
     61     },
     62     'Array | Matrix': function ArrayMatrix(n) {
     63       return (0, _collection.deepMap)(n, this);
     64     } // TODO: For complex numbers, use the approximation for the Faddeeva function
     65     //  from "More Efficient Computation of the Complex Error Function" (AMS)
     66 
     67   });
     68   /**
     69    * Approximates the error function erf() for x <= 0.46875 using this function:
     70    *               n
     71    * erf(x) = x * sum (p_j * x^(2j)) / (q_j * x^(2j))
     72    *              j=0
     73    */
     74 
     75   function erf1(y) {
     76     var ysq = y * y;
     77     var xnum = P[0][4] * ysq;
     78     var xden = ysq;
     79     var i;
     80 
     81     for (i = 0; i < 3; i += 1) {
     82       xnum = (xnum + P[0][i]) * ysq;
     83       xden = (xden + Q[0][i]) * ysq;
     84     }
     85 
     86     return y * (xnum + P[0][3]) / (xden + Q[0][3]);
     87   }
     88   /**
     89    * Approximates the complement of the error function erfc() for
     90    * 0.46875 <= x <= 4.0 using this function:
     91    *                       n
     92    * erfc(x) = e^(-x^2) * sum (p_j * x^j) / (q_j * x^j)
     93    *                      j=0
     94    */
     95 
     96 
     97   function erfc2(y) {
     98     var xnum = P[1][8] * y;
     99     var xden = y;
    100     var i;
    101 
    102     for (i = 0; i < 7; i += 1) {
    103       xnum = (xnum + P[1][i]) * y;
    104       xden = (xden + Q[1][i]) * y;
    105     }
    106 
    107     var result = (xnum + P[1][7]) / (xden + Q[1][7]);
    108     var ysq = parseInt(y * 16) / 16;
    109     var del = (y - ysq) * (y + ysq);
    110     return Math.exp(-ysq * ysq) * Math.exp(-del) * result;
    111   }
    112   /**
    113    * Approximates the complement of the error function erfc() for x > 4.0 using
    114    * this function:
    115    *
    116    * erfc(x) = (e^(-x^2) / x) * [ 1/sqrt(pi) +
    117    *               n
    118    *    1/(x^2) * sum (p_j * x^(-2j)) / (q_j * x^(-2j)) ]
    119    *              j=0
    120    */
    121 
    122 
    123   function erfc3(y) {
    124     var ysq = 1 / (y * y);
    125     var xnum = P[2][5] * ysq;
    126     var xden = ysq;
    127     var i;
    128 
    129     for (i = 0; i < 4; i += 1) {
    130       xnum = (xnum + P[2][i]) * ysq;
    131       xden = (xden + Q[2][i]) * ysq;
    132     }
    133 
    134     var result = ysq * (xnum + P[2][4]) / (xden + Q[2][4]);
    135     result = (SQRPI - result) / y;
    136     ysq = parseInt(y * 16) / 16;
    137     var del = (y - ysq) * (y + ysq);
    138     return Math.exp(-ysq * ysq) * Math.exp(-del) * result;
    139   }
    140 });
    141 /**
    142  * Upper bound for the first approximation interval, 0 <= x <= THRESH
    143  * @constant
    144  */
    145 
    146 exports.createErf = createErf;
    147 var THRESH = 0.46875;
    148 /**
    149  * Constant used by W. J. Cody's Fortran77 implementation to denote sqrt(pi)
    150  * @constant
    151  */
    152 
    153 var SQRPI = 5.6418958354775628695e-1;
    154 /**
    155  * Coefficients for each term of the numerator sum (p_j) for each approximation
    156  * interval (see W. J. Cody's paper for more details)
    157  * @constant
    158  */
    159 
    160 var P = [[3.16112374387056560e00, 1.13864154151050156e02, 3.77485237685302021e02, 3.20937758913846947e03, 1.85777706184603153e-1], [5.64188496988670089e-1, 8.88314979438837594e00, 6.61191906371416295e01, 2.98635138197400131e02, 8.81952221241769090e02, 1.71204761263407058e03, 2.05107837782607147e03, 1.23033935479799725e03, 2.15311535474403846e-8], [3.05326634961232344e-1, 3.60344899949804439e-1, 1.25781726111229246e-1, 1.60837851487422766e-2, 6.58749161529837803e-4, 1.63153871373020978e-2]];
    161 /**
    162  * Coefficients for each term of the denominator sum (q_j) for each approximation
    163  * interval (see W. J. Cody's paper for more details)
    164  * @constant
    165  */
    166 
    167 var Q = [[2.36012909523441209e01, 2.44024637934444173e02, 1.28261652607737228e03, 2.84423683343917062e03], [1.57449261107098347e01, 1.17693950891312499e02, 5.37181101862009858e02, 1.62138957456669019e03, 3.29079923573345963e03, 4.36261909014324716e03, 3.43936767414372164e03, 1.23033935480374942e03], [2.56852019228982242e00, 1.87295284992346047e00, 5.27905102951428412e-1, 6.05183413124413191e-2, 2.33520497626869185e-3]];
    168 /**
    169  * Maximum/minimum safe numbers to input to erf() (in ES6+, this number is
    170  * Number.[MAX|MIN]_SAFE_INTEGER). erf() for all numbers beyond this limit will
    171  * return 1
    172  */
    173 
    174 var MAX_NUM = Math.pow(2, 53);