Convolution

Perform convolution operations on one dimensional arrays. This library provides a flexible direct convolution implementation. The library supports both regular arrays and typed arrays.

Note that this library does not provide a fast (Fast Fourier transform based) convolution implementation. If you are looking for a fast convolution implementation, take a look at ndarray-convolve.

Installation

 $ npm i convolution

Usage

Convolve

convolve(a: Array<T> | TypedArray, b: Array<T> | TypedArray): Array<T> | TypedArray

Performs a convolution operation on two arrays one dimensional a and b. The function returns a new array of the same type as the input, which represents the result of the convolution.

import convolve from "convolution"

const a = [1, 2, 3]
const b = [1, 2, 3]

const result = convolve(a, b)
// result = [1, 4, 10, 12, 9]

Convolve Arbitrary

convolveArbitrary(convolutionStepFunction: (a: List, b: List) => T): (a: List, b: List) => List

Creates a convolution function given a custom convolution step function. A convolution step is the operation of combining two arrays of the same length into a single value. Typically this is done by $\sum_{i}^{}(a_i * b_i)$, but can be arbitrary. The function returns a new function which can be used to perform the whole convolution on two arrays.

The following example is the same as the convolve function above.

import { convolveArbitrary } from "convolution"

const convolutionStepFunction = (a, b) => {
  let sum = 0
  for (let i = 0; i < a.length; i++) {
    sum += a[i] * b[i]
  }
  return sum
}

const convolve = convolveArbitrary(convolutionStepFunction)

const a = [1, 2, 3]
const b = [1, 2, 3]

const result = convolve(a, b)
// result = [1, 4, 10, 12, 9]

Supported Typed Arrays

Array
Int8Array
Uint8Array
Int16Array
Uint16Array
Int32Array
Uint32Array
Float32Array
Float64Array

Contribute

All feedback is appreciated. Create a pull request or write an issue.