Stability and Convergence of Stochastic Gradient Clipping:
            Beyond Lipschitz Continuity and Smoothness

                       Vien V. Mai 1 Mikael Johansson 1

             Abstract                problems are at the core of many machine-learning appli-
                                 cations, and are often solved using stochastic (sub)gradient
    Stochastic gradient algorithms are often unsta-      methods. In spite of their successes, stochastic gradient
    ble when applied to f ...