#laplace

Question 1: Single choice

A regression network models the target given $x$ as a Laplace distribution with location $\mu(x)$ and a fixed scale $s$ , density proportional to $\exp(-|y - \mu|/s)$ . Minimizing the NLL is equivalent to minimizing which loss, and what does $\mu(x)$ then estimate?

Absolute error; the conditional median1
Squared error; the conditional mean2
Absolute error; the conditional mean3
Squared error; the conditional median4

Timed test