Optimizing and certifying bounds of random functions over the hypercube

We consider the problem of certifying an upper bound on the maximum value of a random quadratic form over the hypercube, which corresponds to the problem of optimizing the Hamiltonian of the Sherrington-Kirkpatrick model of statistical physics. We will show that, conditional on the “low-degree polynomials conjecture” concerning the computational hardness of random problems, there is no polynomial-time algorithm certifying a better upper bound than the largest eigenvalue of the coefficient matrix. If time permits we will discuss connections to optimization in random graphs.