= Unbiased stochastic gradient
A random <vector> $G_k$ is an <unbiased stochastic gradient> at an iterate $x_k$ if
$$
\mathbb E[G_k\mid x_k]=\nabla f(x_k).
$$
For an adaptive iteration it is often useful to require the stronger historical condition $\mathbb E[G_k\mid\mathcal F_k]=\nabla f(x_k)$, where $\mathcal F_k$ contains all information available before drawing the new sample. This states explicitly which information is held fixed in the <conditional expectation>.
For $f=m^{-1}\sum_i f_i$, a conditionally uniform index $I_k$ gives $G_k=\nabla f_{I_k}(x_k)$. If index $i$ instead has positive sampling probability $p_i$, then $\nabla f_i(x_k)/(mp_i)$ is the correctly weighted estimate. Omitting these factors generally changes its <expectation>.
For example, average $f_1(x)=(x-1)^2/2$ and $f_2(x)=(x+1)^2/2$. The two sampled gradients are $x-1$ and $x+1$, whose average is the true <gradient> $x$. At the minimizer zero, each sample is nonzero even though its mean vanishes; unbiasedness does not mean zero <variance>.
Back to article page