ntopng/httpdocs/js/ASAP.js
2018-07-30 18:54:52 +02:00

418 lines
14 KiB
JavaScript

/*
* Free FFT and convolution (JavaScript)
*
* Copyright (c) 2014 Project Nayuki
* https://www.nayuki.io/page/free-small-fft-in-multiple-languages
*
* (MIT License)
* Permission is hereby granted, free of charge, to any person obtaining a copy of
* this software and associated documentation files (the "Software"), to deal in
* the Software without restriction, including without limitation the rights to
* use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
* the Software, and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
* - The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
* - The Software is provided "as is", without warranty of any kind, express or
* implied, including but not limited to the warranties of merchantability,
* fitness for a particular purpose and noninfringement. In no event shall the
* authors or copyright holders be liable for any claim, damages or other
* liability, whether in an action of contract, tort or otherwise, arising from,
* out of or in connection with the Software or the use or other dealings in the
* Software.
*/
"use strict";
/* https://github.com/stanford-futuredata/ASAP/blob/master/ASAP-optimized.js */
/*
* Computes the discrete Fourier transform (DFT) of the given complex vector, storing the result back into the vector.
* The vector can have any length. This is a wrapper function.
*/
function transform(real, imag) {
if (real.length != imag.length)
throw "Mismatched lengths";
var n = real.length;
if (n == 0)
return;
else if ((n & (n - 1)) == 0) // Is power of 2
transformRadix2(real, imag);
else // More complicated algorithm for arbitrary sizes
transformBluestein(real, imag);
}
/*
* Computes the inverse discrete Fourier transform (IDFT) of the given complex vector, storing the result back into the vector.
* The vector can have any length. This is a wrapper function. This transform does not perform scaling, so the inverse is not a true inverse.
*/
function inverseTransform(real, imag) {
transform(imag, real);
}
/*
* Computes the discrete Fourier transform (DFT) of the given complex vector, storing the result back into the vector.
* The vector's length must be a power of 2. Uses the Cooley-Tukey decimation-in-time radix-2 algorithm.
*/
function transformRadix2(real, imag) {
// Initialization
if (real.length != imag.length)
throw "Mismatched lengths";
var n = real.length;
if (n == 1) // Trivial transform
return;
var levels = -1;
for (var i = 0; i < 32; i++) {
if (1 << i == n)
levels = i; // Equal to log2(n)
}
if (levels == -1)
throw "Length is not a power of 2";
var cosTable = new Array(n / 2);
var sinTable = new Array(n / 2);
for (var i = 0; i < n / 2; i++) {
cosTable[i] = Math.cos(2 * Math.PI * i / n);
sinTable[i] = Math.sin(2 * Math.PI * i / n);
}
// Bit-reversed addressing permutation
for (var i = 0; i < n; i++) {
var j = reverseBits(i, levels);
if (j > i) {
var temp = real[i];
real[i] = real[j];
real[j] = temp;
temp = imag[i];
imag[i] = imag[j];
imag[j] = temp;
}
}
// Cooley-Tukey decimation-in-time radix-2 FFT
for (var size = 2; size <= n; size *= 2) {
var halfsize = size / 2;
var tablestep = n / size;
for (var i = 0; i < n; i += size) {
for (var j = i, k = 0; j < i + halfsize; j++, k += tablestep) {
var tpre = real[j+halfsize] * cosTable[k] + imag[j+halfsize] * sinTable[k];
var tpim = -real[j+halfsize] * sinTable[k] + imag[j+halfsize] * cosTable[k];
real[j + halfsize] = real[j] - tpre;
imag[j + halfsize] = imag[j] - tpim;
real[j] += tpre;
imag[j] += tpim;
}
}
}
// Returns the integer whose value is the reverse of the lowest 'bits' bits of the integer 'x'.
function reverseBits(x, bits) {
var y = 0;
for (var i = 0; i < bits; i++) {
y = (y << 1) | (x & 1);
x >>>= 1;
}
return y;
}
}
/*
* Computes the discrete Fourier transform (DFT) of the given complex vector, storing the result back into the vector.
* The vector can have any length. This requires the convolution function, which in turn requires the radix-2 FFT function.
* Uses Bluestein's chirp z-transform algorithm.
*/
function transformBluestein(real, imag) {
// Find a power-of-2 convolution length m such that m >= n * 2 + 1
if (real.length != imag.length)
throw "Mismatched lengths";
var n = real.length;
var m = 1;
while (m < n * 2 + 1)
m *= 2;
// Trignometric tables
var cosTable = new Array(n);
var sinTable = new Array(n);
for (var i = 0; i < n; i++) {
var j = i * i % (n * 2); // This is more accurate than j = i * i
cosTable[i] = Math.cos(Math.PI * j / n);
sinTable[i] = Math.sin(Math.PI * j / n);
}
// Temporary vectors and preprocessing
var areal = new Array(m);
var aimag = new Array(m);
for (var i = 0; i < n; i++) {
areal[i] = real[i] * cosTable[i] + imag[i] * sinTable[i];
aimag[i] = -real[i] * sinTable[i] + imag[i] * cosTable[i];
}
for (var i = n; i < m; i++)
areal[i] = aimag[i] = 0;
var breal = new Array(m);
var bimag = new Array(m);
breal[0] = cosTable[0];
bimag[0] = sinTable[0];
for (var i = 1; i < n; i++) {
breal[i] = breal[m - i] = cosTable[i];
bimag[i] = bimag[m - i] = sinTable[i];
}
for (var i = n; i <= m - n; i++)
breal[i] = bimag[i] = 0;
// Convolution
var creal = new Array(m);
var cimag = new Array(m);
convolveComplex(areal, aimag, breal, bimag, creal, cimag);
// Postprocessing
for (var i = 0; i < n; i++) {
real[i] = creal[i] * cosTable[i] + cimag[i] * sinTable[i];
imag[i] = -creal[i] * sinTable[i] + cimag[i] * cosTable[i];
}
}
/*
* Computes the circular convolution of the given real vectors. Each vector's length must be the same.
*/
function convolveReal(x, y, out) {
if (x.length != y.length || x.length != out.length)
throw "Mismatched lengths";
var zeros = new Array(x.length);
for (var i = 0; i < zeros.length; i++)
zeros[i] = 0;
convolveComplex(x, zeros, y, zeros.slice(), out, zeros.slice());
}
/*
* Computes the circular convolution of the given complex vectors. Each vector's length must be the same.
*/
function convolveComplex(xreal, ximag, yreal, yimag, outreal, outimag) {
if (xreal.length != ximag.length || xreal.length != yreal.length || yreal.length != yimag.length || xreal.length != outreal.length || outreal.length != outimag.length)
throw "Mismatched lengths";
var n = xreal.length;
xreal = xreal.slice();
ximag = ximag.slice();
yreal = yreal.slice();
yimag = yimag.slice();
transform(xreal, ximag);
transform(yreal, yimag);
for (var i = 0; i < n; i++) {
var temp = xreal[i] * yreal[i] - ximag[i] * yimag[i];
ximag[i] = ximag[i] * yreal[i] + xreal[i] * yimag[i];
xreal[i] = temp;
}
inverseTransform(xreal, ximag);
for (var i = 0; i < n; i++) { // Scaling (because this FFT implementation omits it)
outreal[i] = xreal[i] / n;
outimag[i] = ximag[i] / n;
}
}
/**********************************************************************************/
function binarySearch(head, tail, data, minObj, originalKurt, windowSize) {
while (head <= tail) {
var w = Math.round((head + tail) / 2);
var smoothed = SMA(data, w, 1);
var metrics = new Metrics(smoothed);
if (metrics.kurtosis() >= originalKurt) { /* Search second half if feasible */
var roughness = metrics.roughness();
if (roughness < minObj) {
windowSize = w;
minObj = roughness;
}
head = w + 1;
} else { /* Search first half */
tail = w - 1;
}
}
return windowSize;
}
function smooth(data, resolution) {
/* Ignore the last value if it's NaN */
if (isNaN(data[data.length-1])) {
data = data.slice(0, -1);
}
if (data.length >= 2 * resolution ) {
data = SMA(data, Math.trunc(data.length / resolution),
Math.trunc(data.length / resolution));
}
var acf = new ACF(data, Math.round(data.length / 10));
var peaks = acf.findPeaks();
var metrics = new Metrics(data);
var originalKurt = metrics.kurtosis();
var minObj = metrics.roughness();
var windowSize = 1;
var lb = 1;
var largestFeasible = -1;
var tail = data.length / 10;
for (var i = peaks.length - 1; i >= 0; i -=1) {
var w = peaks[i];
if (w < lb || w == 1) {
break;
} else if (Math.sqrt(1 - acf.correlations[w]) * windowSize >
Math.sqrt(1 - acf.correlations[windowSize]) * w) {
continue;
}
var smoothed = SMA(data, w, 1);
metrics = new Metrics(smoothed);
var roughness = metrics.roughness();
if (metrics.kurtosis() >= originalKurt) {
if (roughness < minObj) {
minObj = roughness;
windowSize = w;
}
lb = Math.round(Math.max(w * Math.sqrt((acf.maxACF - 1) / (acf.correlations[w] - 1)), lb));
if (largestFeasible < 0) { largestFeasible = i; }
}
}
if (largestFeasible > 0) {
if (largestFeasible < peaks.length - 2) { tail = peaks[largestFeasible + 1]; }
lb = Math.max(lb, peaks[largestFeasible] + 1);
}
windowSize = binarySearch(lb, tail, data, minObj, originalKurt, windowSize);
return SMA(data, windowSize, 1);
}
function SMA(data, range, slide) {
var windowStart = 0;
var sum = 0;
var count = 0;
var values = [];
for (var i = 0; i < data.length; i ++) {
if (isNaN(data[i])) { data[i] = 0; }
if (i - windowStart >= range) {
values.push(sum / count);
var oldStart = windowStart;
while (windowStart < data.length && windowStart - oldStart < slide) {
sum -= data[windowStart];
count -= 1;
windowStart += 1;
}
}
sum += data[i];
count += 1;
}
if (count == range) {
values.push(sum / count);
}
return values;
}
function ACF(values, maxLag) {
this.mean = Metrics.mean(values);
this.values = values;
this.correlations = new Array(maxLag);
this.CORR_THRESH = 0.2;
this.maxACF = 0;
this.calculate();
}
ACF.prototype.calculate = function() {
/* Padding to the closest power of 2 */
var len = Math.pow(2, Math.trunc(Math.log2(this.values.length)) + 1)
var fftreal = new Array(len).fill(0);
var fftimg = new Array(len).fill(0);
for (var i = 0; i < this.values.length; i += 1) {
fftreal[i] = this.values[i] - this.mean;
}
/* F_R(f) = FFT(X) */
transform(fftreal, fftimg);
/* S(f) = F_R(f)F_R*(f) */
for (var i = 0; i < fftreal.length; i += 1) {
fftreal[i] = Math.pow(fftreal[i], 2) + Math.pow(fftimg[i], 2);
fftimg[i] = 0;
}
/* R(t) = IFFT(S(f)) */
inverseTransform(fftreal, fftimg);
for (var i = 1; i < this.correlations.length; i += 1) {
this.correlations[i] = fftreal[i] / fftreal[0];
}
}
ACF.prototype.findPeaks = function() {
var peakIndices = [];
if (this.correlations.length > 1) {
var positive = this.correlations[1] > this.correlations[0];
var max = 1;
for (var i = 2; i < this.correlations.length; i += 1) {
if (!positive && this.correlations[i] > this.correlations[i - 1]) {
max = i;
positive = !positive;
} else if (positive && this.correlations[i] > this.correlations[max]) {
max = i;
} else if (positive && this.correlations[i] < this.correlations[i - 1]) {
if (max > 1 && this.correlations[max] > this.CORR_THRESH) {
peakIndices.push(max);
if (this.correlations[max] > this.maxACF) {
this.maxACF = this.correlations[max];
}
}
positive = !positive;
}
}
}
/* If there is no autocorrelation peak within the MAX_WINDOW boundary,
# try windows from the largest to the smallest */
if (peakIndices.length <= 1) {
for (var i = 2; i < this.correlations.length; i += 1) {
peakIndices.push(i);
}
}
return peakIndices;
}
function Metrics(values) {
this.len = values.length;
this.values = values;
this.m = Metrics.mean(values);
}
Metrics.mean = function(values) {
var m = 0;
for (var i = 0; i < values.length; i += 1) {
m += values[i];
}
return m / values.length;
}
Metrics.std = function(values) {
var m = Metrics.mean(values);
var std = 0;
for (var i = 0; i < values.length; i += 1) {
std += Math.pow((values[i] - m), 2);
}
return Math.sqrt(std / values.length);
}
Metrics.prototype.kurtosis = function() {
var u4 = 0, variance = 0;
for (var i = 0; i < this.len; i ++) {
u4 += Math.pow((this.values[i] - this.m), 4);
variance += Math.pow((this.values[i] - this.m), 2);
}
return this.len * u4 / Math.pow(variance, 2);
}
Metrics.prototype.roughness = function() {
return Metrics.std(this.diffs());
}
Metrics.prototype.diffs = function() {
var diff = new Array(this.len - 1);
for (var i = 1; i < this.len; i += 1) {
diff[i - 1] = this.values[i] - this.values[i - 1];
}
return diff;
}