Saddle-point problems have recently gained increased attention from the machine learning community, mainly due to applications in training Generative Adversarial Networks using stochastic gradients. At same time, some only a zeroth-order oracle is available. In this paper, we propose several algorithms solve smooth (strongly) convex-concave saddle-point oracles and estimate their convergence ra...

First order logic without quantifiers, a.k.a. zeroth order logic, is more expressive than propositional logic. In this paper, we introduce two new operators for regular expressions allowing us to handle zeroth order boolean formula in order to increase the expressive power of expressions; We also extend the notion of interpretation and of realization using expression interpretation. We thus def...

In this paper, we consider a stochastic distributed nonconvex optimization problem with the cost function being over n agents having access only to zeroth-order (ZO) information of cost. This has various machine learning applications. As solution, propose two ZO algorithms, in which at each iteration agent samples local oracle points time-varying smoothing parameter. We show that proposed algor...

We consider stochastic zeroth-order optimization over Riemannian submanifolds embedded in Euclidean space, where the task is to solve problems with only noisy objective function evaluations. Toward this, our main contribution propose estimators of gradient and Hessian from evaluations, based on a version Gaussian smoothing technique. The proposed overcome difficulty nonlinearity manifold constr...

We consider the problem of minimizing a smooth, Lipschitz, convex function over compact, set using sub-zeroth-order oracles: an oracle that outputs sign directional derivative for given point and direction, compares values pair points, noisy value point. show sample complexity optimization these oracles is polynomial in relevant parameters. The algorithm we provide comparator first with known r...

In this paper, we show that in the class of connected graphs G of order n ≥ 3 having girth at least equal to k, 3 ≤ k ≤ n, the unique graph G having minimum general sum-connectivity index χα(G) consists of Ck and n−k pendant vertices adjacent to a unique vertex of Ck, if−1 ≤ α < 0. This property does not hold for zeroth-order general Randić index Rα(G).

In this paper, we propose a class of penalty methods with stochastic approximation for solving stochastic nonlinear programming problems. We assume that only noisy gradients or function values of the objective function are available via calls to a stochastic first-order or zeroth-order oracle. In each iteration of the proposed methods, we minimize an exact penalty function which is nonsmooth an...