Skew-symmetrically perturbed gradient flow for convex optimization