First assume $\zeta_i=0$, which conditions guarantee that $w^*$ is a stable fixed point? If it's not a stable fixed point for noise-free case, then the process will not be ergodic with additive noise. Consider \begin{equation} \Sigma=\left( \begin{matrix} 1 & 0 \\ 0 & 2 \end{matrix} \right) \end{equation} One condition on step size $\eta=\alpha$ which guarantees convergence $$\frac{2}{\alpha}<\text{Tr}(\Sigma)+2\|H\|$$ Which gives that $$\alpha<\frac{2}{7}$$ However, this condition is too strict. The necessary and sufficient condition is that $$\alpha<\frac {4} {9 + \sqrt {17}}$$ This [note](http://yaroslavvb.com/convergence.pdf) derives necessary and sufficient condition above for a general Gaussian. I haven't seen anyone else derives this, please correct me if it occurs in literature. For general Gaussian, the necessary and sufficient condition for $w^*$ to be a fixed point is that $$\alpha<\frac{2}{\rho(A)}$$ Here, $\rho$ to denotes spectral radius and $A$ is defined below. Let $s=(s_1, s_2, s_3, \ldots)$ be the eigenvalues of $\Sigma$, then \begin{equation}\label{defa} A= 2 \left( \begin{array}{cccc} s_1 & 0 & 0 & \ldots \\ 0 & s_2 & 0 & \ldots \\ 0 & 0 & s_3 & \ldots \\ \ldots & \ldots & \ldots & \ldots \end{array} \right) + \left( \begin{array}{cccc} s_1 & s_1 & s_1 & \ldots \\ s_2 & s_2 & s_2 & \ldots \\ s_3 & s_3 & s_3 & \ldots \\ \ldots & \ldots & \ldots & \ldots \end{array} \right) \end{equation} Perhaps there's an easier expression that doesn't require forming matrix $A$? --- Background for non-Gaussian case: It can be shown that in the case of 1 dimension and deterministic $x$, the following condition on $\alpha$ is necessary and sufficient for convergence \begin{equation}\label{supersimple} \alpha x^4 < 2 x^2 \end{equation} Since we have $h=x^2$ for Hessian $h$, this reduces to the well known bound on convergent learning rate: $\alpha < 2/h$ In the case of stochastic x, the following is necessary and sufficient \begin{equation}\label{eq:0} \alpha E[x^4] < 2 E[x^2] \end{equation} For the case of $x$ being distributed as standard normal, this gives $2/(3h)$ for the largest learning rate, three times smaller than what's allowed in deterministic case For the case of $d$ dimensions, the following is a sufficient condition, with $\prec$ indicating Loewner order\footnote{assumption A.6 in Bach [paper](https://www.di.ens.fr/~fbach/aistats_defossez_bach_with_supp.pdf) \begin{equation}\label{eq:1} \alpha E[xx'xx'] \prec E[xx'] \end{equation} The right-hand side can be tightened to \begin{equation}\label{eq:1x} \alpha E[xx'xx'] \prec 2 E[xx'] \end{equation} Defossez, Deffosez2015 [showed](https://www.di.ens.fr/~fbach/aistats_defossez_bach_with_supp.pdf) that the following optimization over symmetric matrices gives sufficient condition for convergence, and conjectured it to also be necessary (Lemma 1 of Defossez2015) \begin{equation}\label{eq:2} \frac{1}{\alpha} < \sup_{A\in \mathcal{S}(R^d)} \frac{E[(x'Ax)^2]}{2 E[x'A^2 x] } \end{equation} We can show this to be equivalent to the following positive semi-definite constraint \begin{equation}\label{eq:3} \alpha E[xx' \otimes xx'] \prec E[xx'\otimes I] + E[I\otimes xx'] \end{equation} Most recently, [Jain](https://arxiv.org/abs/1610.03774) generalized last Eq to batch sizes beyond 1 and formally showed it to be a necessary condition for convergence.