We present a general theory of exact penalty functions with vectorial (multidimensional) penalty parameter for optimization problems in infinite dimensional spaces. In comparison with the scalar case, the use of vectorial penalty parameters provides much more flexibility, allows one to adaptively and independently take into account the violation of each constraint during an optimization process, and often leads to a better overall performance of an optimization method using an exact penalty function. We obtain sufficient conditions for the local and global exactness of penalty functions with vectorial penalty parameters and study convergence of global exact penalty methods with several different penalty updating strategies. In particular, we present a new algorithmic approach to an analysis of the global exactness of penalty functions, which contains a novel characterisation of the global exactness property in terms of behaviour of sequences generated by certain optimization methods.
This paper focuses on the design of sequential quadratic optimization (commonly known as SQP) methods for solving large-scale nonlinear optimization problems. The most computationally demanding aspect of such an approach is the computation of the search direction during each iteration, for which we consider the use of matrix-free methods. In particular, we develop a method that requires an inexact solve of a single QP subproblem to establish the convergence of the overall SQP method. It is known that SQP methods can be plagued by poor behavior of the global convergence mechanism. To confront t…
A Fixed-Penalty Linearized Augmented Lagrangian Method with Classical Multiplier Updates
Beijing International Center for Mathematical Research, Peking University（北京大学北京国际数学研究中心）；College of Science, National University of Defense Technology（国防科技大学理学院）；School of Mathematical Sciences, Peking University（北京大学数学科学学院）
A penalty-type method for relaxed inverse optimal control problems
University of Siegen（锡根大学）；Higher Training College, University of Yaoundé I（雅温得第一大学高等培训学院）；AIMS-Cameroon Research Center（AIMS喀麦隆研究中心）；University of Marburg（马尔堡大学）
We present a parallelized primal-dual algorithm for solving constrained convex optimization problems. The algorithm is "block-based," in that vectors of primal and dual variables are partitioned into blocks, each of which is updated only by a single processor. We consider four possible forms of asynchrony: in updates to primal variables, updates to dual variables, communications of primal variables, and communications of dual variables. We explicitly construct a family of counterexamples to rule out permitting asynchronous communication of dual variables, though the other forms of asynchrony a…
We present a parallelized primal-dual algorithm for solving constrained convex optimization problems. The algorithm is "block-based," in that vectors of primal and dual variables are partitioned into blocks, each of which is updated only by a single processor. We consider four possible forms of asynchrony: in updates to primal variables, updates to dual variables, communications of primal variables, and communications of dual variables. We construct a family of explicit counterexamples to show the need to eliminate asynchronous communication of dual variables, though the other forms of asynchr…
Simultaneous perturbation stochastic approximation (SPSA) is widely used in stochastic optimization due to its high efficiency, asymptotic stability, and reduced number of required loss function measurements. However, the standard SPSA algorithm needs to be modified to deal with constrained problems. In recent years, sequential quadratic programming (SQP)-based projection ideas and penalty ideas have been analyzed. Both ideas have convergence results and a potentially wide range of applications, but with some limitations in practical consideration, such as computation time, complexity, and fea…
This paper proposes and analyzes a dampened proximal alternating direction method of multipliers (DP.ADMM) for solving linearly-constrained nonconvex optimization problems where the smooth part of the objective function is nonseparable. Each iteration of DP.ADMM consists of: (i) a sequence of partial proximal augmented Lagrangian (AL) updates, (ii) an under-relaxed Lagrange multiplier update, and (iii) a novel test to check whether the penalty parameter of the AL function should be updated. Under a basic Slater condition and some requirements related to the dampening factor and under-relaxatio…
We implement the adaptive step size scheme from the optimization methods AdaGrad and Adam in a novel variant of the Proximal Gradient Method (PGM). Our algorithm, dubbed AdaProx, avoids the need for explicit computation of the Lipschitz constants or additional line searches and thus reduces per-iteration cost. In test cases for Constrained Matrix Factorization we demonstrate the advantages of AdaProx in fidelity and performance over PGM, while still allowing for arbitrary penalty functions. The python implementation of the algorithm presented here is available as an open-source package at http…
Fast and Effective Computation of Generalized Symmetric Matrix Factorization
School of Computer Science and Engineering（计算机科学与工程学院）；Guangdong Province Key Laboratory of Computational Science（广东省计算科学重点实验室）；Sun Yat-sen University（中山大学）；School of Mathematics and Information Science（数学与信息科学学院）；Guangzhou University（广州大学）；Department of Mathematics（数学系）；The University of Tennessee, Knoxville（田纳西大学，诺克斯维尔）
Tsinghua Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院）；JD Explore Academy（京东探索研究院）；School of Artificial Intelligence, Peking University（北京大学人工智能学院）
Safe reinforcement learning aims to learn the optimal policy while satisfying safety constraints, which is essential in real-world applications. However, current algorithms still struggle for efficient policy updates with hard constraint satisfaction. In this paper, we propose Penalized Proximal Policy Optimization (P3O), which solves the cumbersome constrained policy iteration via a single minimization of an equivalent unconstrained problem. Specifically, P3O utilizes a simple-yet-effective penalty function to eliminate cost constraints and removes the trust-region constraint by the clipped s…
