gaussian-fit

Fit time/spectrum/other sequential data with a set of gaussians by expectation-maximization algoritm.

Usage

const fit = require('gaussian-fit')

//time/frequency/etc series
let data = [0, .1, .2, .5, .2, .1, 0]

let how = {
	//predefined components, each one is object {weight, mean, variance}
	//if null - the components will be detected automatically
	components: null,

	//max number of components in case of auto-detection
	maxNumber: 100,

	//max number of iterations
	maxIterations: 200,

	//min difference of likelihood
	tolerance: 1e-5,
}

let components = fit(data, how)

components.forEach({weight, mean, variance} => {
	//every component is an object with weight, mean and variance properties
})

fit(data, components?)

Also see webgl version gaussian-fit/gl for better performance.

Similar

• gaussianmixture — classical gaussian mixture for 1d samples