= Median minimizes expected absolute loss
{title2=$\arg\min_a\mathbb E|X-a|$}
For a finite <absolute moment>[first absolute moment], the minimizers of <absolute loss>[expected absolute loss] are the <medians>. With a <probability density function>[density], differentiating the loss gives $2F(a)-1$, so every point where the <cumulative distribution function> equals one half minimizes it. A <probability density function>[density] gap may give a whole minimizing <interval>. In contrast, finite squared loss is uniquely minimized by the <expectation>.
Back to article page