multiscale-array

Create multiple scales representation for an array (see scale-space for the concept). It is simplest form of convolution with resampling, where the upper level is perfectly 2 times less than the lower level, with each value made up of two values from the previous level. Similar to mipmaps, but 1d.

const multiscale = require('multiscale-array')
 
let data = Array(1e7)
let scales = multiscale(data, {
    //max group of samples
    maxScale: 65536,
 
    //how to form upper level from two samples of lower level
    reduce: (left, right, i, level) => .5*left + .5*right
})
 
//recalculate scales for the data range, this is O(2N)
scales.update(from?, to?)
 
//replace old data with the new data and recalc scales for it
scales.update(newData, from?, to?)
 
//get data for the scale 2⁴
let scaleData = scales[4];
 
//subset scales, i.e. slice all scales, but mutable (!)
scales.subset(from, to)