Developer Reference for Intel® oneAPI Math Kernel Library for C

ID 766684
Date 12/16/2022
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

Convolution and Correlation Mathematical Notation and Definitions

The following notation is necessary to explain the underlying mathematical definitions used in the text:

R = (-, +) The set of real numbers.
Z = {0, ±1, ±2, ...} The set of integer numbers.
ZN = Z× ... ×Z The set of N-dimensional series of integer numbers.
p = (p1, ..., pN) ZN N-dimensional series of integers.
u:ZNR Function u with arguments from ZN and values from R.
u(p) = u(p1, ..., pN) The value of the function u for the argument (p1, ..., pN).
w = u*v Function w is the convolution of the functions u, v.
w = uv Function w is the correlation of the functions u, v.

Given series p, qZN:

  • series r = p + q is defined as rn = pn + qn for every n=1,...,N

  • series r = p - q is defined as rn = pn - qn for every n=1,...,N

  • series r = sup{p, q} is defines as rn = max{pn, qn} for every n=1,...,N

  • series r = inf{p, q} is defined as rn = min{pn, qn} for every n=1,...,N

  • inequality pq means that pnqn for every n=1,...,N.

A function u(p) is called a finite function if there exist series Pmin, PmaxZN such that:

u(p) 
         0 
      
implies
 Pminp Pmax.

Operations of convolution and correlation are only defined for finite functions.

Consider functions u, v and series Pmin, PmaxQmin, QmaxZN such that:

u(p) 0 implies Pminp Pmax.
v(q) 0 implies Qminq Qmax.

Definitions of linear correlation and linear convolution for functions u and v are given below.

Linear Convolution

If function w = u*v is the convolution of u and v, then:

w(r) 0 implies RminrRmax,
where Rmin = Pmin + Qmin and Rmax = Pmax + Qmax.

If RminrRmax, then:

w(r) = u(tv(rt) is the sum for all tZN such that TmintTmax,
where Tmin = sup{Pmin, r Qmax} and Tmax = inf{Pmax, r Qmin}.

Linear Correlation

If function w = uv is the correlation of u and v, then:

w(r) 0 implies RminrRmax,
where Rmin = Qmin - Pmax and Rmax = Qmax - Pmin.

If RminrRmax, then:

w(r) = u(tv(r+t) is the sum for all tZN such that TmintTmax,
where Tmin = sup{Pmin, Qminr} and Tmax = inf{Pmax, Qmaxr}.

Representation of the functions u, v, was the input/output data for the Intel® oneAPI Math Kernel Library convolution and correlation functions is described in theData Allocation.