Canonical: https://knowlpedia.com/measure-theory/pushforward-measure/

# Pushforward measure

The measure obtained by transporting a measure through a measurable map.

A **pushforward measure** transports a measure along a measurable map. Let \((X,\mathcal A,\mu)\) be a [measure space](https://knowlpedia.com/measure-theory/measure-space/), let \((Y,\mathcal B)\) be a [measurable space](https://knowlpedia.com/measure-theory/measurable-space/), and let \(T:X\to Y\) be a [measurable function](https://knowlpedia.com/measure-theory/measurable-function/). The **pushforward** of \(\mu\) by \(T\), denoted \(T_\#\mu\) or \(T_*\mu\), is the [measure](https://knowlpedia.com/measure-theory/measure/) on \((Y,\mathcal B)\) defined by
\[
(T_\#\mu)(B)=\mu(T^{-1}(B))\qquad(B\in\mathcal B).
\]

The definition measures subsets of \(Y\) by pulling them back to \(X\). Pushforwards are the natural language for the [change-of-variables formula](https://knowlpedia.com/measure-theory/change-of-variables-pushforward/).

## Examples

- Let \(\lambda\) be Lebesgue measure on \([0,1]\) and let \(T(x)=x^2\). Then \(\nu=T_\#\lambda\) satisfies \(\nu([0,t])=\sqrt t\) for \(0\le t\le1\), so \(\nu\) has density \(1/(2\sqrt y)\) on \((0,1]\).
- If \(\pi_X:X\times Y\to X\) is projection and \(\mu\otimes\nu\) is a [product measure](https://knowlpedia.com/measure-theory/product-measure/) with \(\nu(Y)<\infty\), then
  \[
  (\pi_X)_\#(\mu\otimes\nu)(A)=(\mu\otimes\nu)(A\times Y)=\mu(A)\nu(Y)
  \]
  for \(A\in\mathcal A\). Thus \((\pi_X)_\#(\mu\otimes\nu)=\nu(Y)\mu\); in particular, it equals \(\mu\) when \(\nu\) is a probability measure.
