Showing 1–3 of 3 results for author: Pichugin, A

Method with Batching for Stochastic Finite-Sum Variational Inequalities in Non-Euclidean Setting

Authors: Alexander Pichugin, Maksim Pechin, Aleksandr Beznosikov, Vasilii Novitskii, Alexander Gasnikov

Abstract: Variational inequalities are a universal optimization paradigm that incorporate classical minimization and saddle point problems. Nowadays more and more tasks require to consider stochastic formulations of optimization problems. In this paper, we present an analysis of a method that gives optimal convergence estimates for monotone stochastic finite-sum variational inequalities. In contrast to the… ▽ More Variational inequalities are a universal optimization paradigm that incorporate classical minimization and saddle point problems. Nowadays more and more tasks require to consider stochastic formulations of optimization problems. In this paper, we present an analysis of a method that gives optimal convergence estimates for monotone stochastic finite-sum variational inequalities. In contrast to the previous works, our method supports batching, does not lose the oracle complexity optimality and uses an arbitrary Bregman distance to take into account geometry of the problem. Paper provides experimental confirmation to algorithm's effectiveness. △ Less

Submitted 15 September, 2024; v1 submitted 13 August, 2024; originally announced August 2024.

Comments: 38 pages, 1 algorithm, 4 figures, 1 table

Optimal Analysis of Method with Batching for Monotone Stochastic Finite-Sum Variational Inequalities

Authors: Alexander Pichugin, Maksim Pechin, Aleksandr Beznosikov, Alexander Gasnikov

Abstract: Variational inequalities are a universal optimization paradigm that is interesting in itself, but also incorporates classical minimization and saddle point problems. Modern realities encourage to consider stochastic formulations of optimization problems. In this paper, we present an analysis of a method that gives optimal convergence estimates for monotone stochastic finite-sum variational inequal… ▽ More Variational inequalities are a universal optimization paradigm that is interesting in itself, but also incorporates classical minimization and saddle point problems. Modern realities encourage to consider stochastic formulations of optimization problems. In this paper, we present an analysis of a method that gives optimal convergence estimates for monotone stochastic finite-sum variational inequalities. In contrast to the previous works, our method supports batching and does not lose the oracle complexity optimality. The effectiveness of the algorithm, especially in the case of small but not single batches is confirmed experimentally. △ Less

Submitted 26 March, 2024; v1 submitted 15 January, 2024; originally announced January 2024.

Comments: 22 pages, 1 algorithm, 2 figures, 1 table

A Statistical Method for Corrupt Agents Detection

Authors: Yury A. Pichugin, Oleg A. Malafeyev, Denis Rylow

Abstract: The statistical method is used to identify the hidden leaders of the corruption structure. The method is based on principal component analysis (PCA), linear regression, and Shannon information. It is applied to study the time series data of corrupt financial activity. Shannon's quantity of information is specified as a function of two arguments: a vector of hidden corruption factors and a subset… ▽ More The statistical method is used to identify the hidden leaders of the corruption structure. The method is based on principal component analysis (PCA), linear regression, and Shannon information. It is applied to study the time series data of corrupt financial activity. Shannon's quantity of information is specified as a function of two arguments: a vector of hidden corruption factors and a subset of corrupt agents. Several optimization problems are solved to determine the contribution of corresponding corrupt agents to the total illegal behavior. An illustrative example is given. A convenient algorithm for computing the covariance matrix with missing data is proposed. △ Less

Submitted 14 July, 2017; originally announced July 2017.