1: \begin{abstract}

2: We consider in this work small random perturbations (of multiplicative noise type) of the gradient flow. We rigorously prove that under mild conditions, when the potential function is a Morse function with additional strong saddle condition, the perturbed gradient flow converges to the neighborhood of local minimizers in $O(\ln (\ve^{-1}))$ time on the average, where $\ve$ is the scale of the random perturbation. Under a change of time scale, this indicates that for the diffusion process that approximates the stochastic gradient method, it takes (up to logarithmic factor) only a linear time of inverse stepsize to evade from all saddle points and hence it implies a fast convergence of its discrete--time counterpart.

3: \end{abstract}

4: