Mathematics

The decentralized solution of linear systems of equations arises as a subproblem in optimization over networks. Typical examples include the KKT system corresponding to equality constrained quadratic programs in distributed optimization algorithms or in active set methods. This note presents a tailored structure-exploiting decentralized variant of the conjugate gradient method. We show that the decentralized conjugate gradient method exhibits super-linear convergence in a finite number of steps. Finally, we illustrate the algorithm's performance in comparison to the Alternating Direction Method of Multipliers drawing upon examples from sensor fusion.

This paper introduces two decomposition-based methods for two-block mixed-integer linear programs (MILPs), which break the original problem into a sequence of smaller MILP subproblems. The first method is based on the l1-augmented Lagrangian. The second method is based on the alternating direction method of multipliers. When the original problem has a block-angular structure, the subproblems of the first block have low dimensions and can be solved in parallel. We add reverse-norm cuts and augmented Lagrangian cuts to the subproblems of the second block. For both methods, we show asymptotic convergence to globally optimal solutions and present iteration upper bounds. Numerical comparisons with recent decomposition methods demonstrate the exactness and efficiency of our proposed methods.

In this work, we study a novel class of projection-based algorithms for linearly constrained problems (LCPs) which have a lot of applications in statistics, optimization, and machine learning. Conventional primal gradient-based methods for LCPs call a projection after each (stochastic) gradient descent, resulting in that the required number of projections equals that of gradient descents (or total iterations). Motivated by the recent progress in distributed optimization, we propose the delayed projection technique that calls a projection once for a while, lowering the projection frequency and improving the projection efficiency. Accordingly, we devise a series of stochastic methods for LCPs using the technique, including a variance reduced method and an accelerated one. We theoretically show that it is feasible to improve projection efficiency in both strongly convex and generally convex cases. Our analysis is simple and unified and can be easily extended to other methods using delayed projections. When applying our new algorithms to federated optimization, a newfangled and privacy-preserving subfield in distributed optimization, we obtain not only a variance reduced federated algorithm with convergence rates better than previous works, but also the first accelerated method able to handle data heterogeneity inherent in federated optimization.

Departure time choice models play a crucial role in determining the traffic load in transportation systems. This paper introduces a new framework to model and analyze the departure time user equilibrium (DTUE) problem based on the so-called Mean Field Games (MFGs) theory. The proposed framework is the combination of two main components including (i) the reaction of travelers to the traffic congestion by choosing their departure times to optimize their travel cost; and (ii) the aggregation of the actions of the travelers, which determines the congestion of the system. The first component corresponds to a classic game theory model while the second one captures the travelers' interactions at the macroscopic level and describes the system dynamics. In this paper, we first present a continuous departure time choice model and investigate the equilibria of the system. Specifically, we demonstrate the existence of the equilibrium and characterize the DTUE. Then, a discrete approximation of the system is provided based on deterministic differential game models to numerically obtain the equilibrium of the system. To examine the efficiency of the proposed model, we compare it with the departure time choice models in the literature. We observe that the solution obtained based on our model is 5.6\% closer to the optimal ones compared to the solutions determined based on models in the literature. Moreover, our proposed model converges much faster with 87\% less number of iterations required to converge. Finally, the model is applied to the real test case of Lyon Metropolis. The results show that the proposed framework is capable of not only considering a large number of players but also including multiple preferred travel times and heterogeneous trip lengths more accurately than existing models in the literature.

There has been work on exploiting polynomial approximation to solve distributed nonconvex optimization problems. This idea facilitates arbitrarily precise global optimization without requiring local evaluations of gradients at every iteration. Nonetheless, there remains a gap between existing theoretical guarantees and diverse practical requirements for dependability, including privacy preservation and robustness to network imperfections (e.g., time-varying directed communication, asynchrony and packet drops). To fill this gap and keep the above strengths, we propose a Dependable Chebyshev-Proxy-based distributed Optimization Algorithm (D-CPOA). Specifically, to ensure both accuracy of solutions and privacy preservation of local objective functions, a new privacy-preserving mechanism is designed. This mechanism leverages the randomness in block-wise insertions of perturbed data and separate subtractions of added noises, and its effects are thoroughly analyzed through ( α,β )-data-privacy. In addition, to gain robustness to various network imperfections, we use the push-sum consensus protocol as a backbone, discuss its specific enhancements, and evaluate the performance of the proposed algorithm accordingly. Thanks to the linear consensus-based structure of iterations, we avoid the privacy-accuracy trade-off and the bother of selecting appropriate step-sizes in different settings. We provide rigorous treatments of the accuracy, dependability and complexity. It is shown that the advantages brought by the idea of polynomial approximation are perfectly maintained when all the above challenging requirements exist. Simulations demonstrate the efficacy of the developed algorithm.

We present a flexible trust region descend algorithm for unconstrained and convexly constrained multiobjective optimization problems. It is targeted at heterogeneous and expensive problems, i.e., problems that have at least one objective function that is computationally expensive. The method is derivative-free in the sense that neither need derivative information be available for the expensive objectives nor are gradients approximated using repeated function evaluations as is the case in finite-difference methods. Instead, a multiobjective trust region approach is used that works similarly to its well-known scalar pendants. Local surrogate models constructed from evaluation data of the true objective functions are employed to compute possible descent directions. In contrast to existing multiobjective trust region algorithms, these surrogates are not polynomial but carefully constructed radial basis function networks. This has the important advantage that the number of data points scales linearly with the parameter space dimension. The local models qualify as fully linear and the corresponding general scalar framework is adapted for problems with multiple objectives. Convergence to Pareto critical points is proven and numerical examples illustrate our findings.

Direct policy search serves as one of the workhorses in modern reinforcement learning (RL), and its applications in continuous control tasks have recently attracted increasing attention. In this work, we investigate the convergence theory of policy gradient (PG) methods for learning the linear risk-sensitive and robust controller. In particular, we develop PG methods that can be implemented in a derivative-free fashion by sampling system trajectories, and establish both global convergence and sample complexity results in the solutions of two fundamental settings in risk-sensitive and robust control: the finite-horizon linear exponential quadratic Gaussian, and the finite-horizon linear-quadratic disturbance attenuation problems. As a by-product, our results also provide the first sample complexity for the global convergence of PG methods on solving zero-sum linear-quadratic dynamic games, a nonconvex-nonconcave minimax optimization problem that serves as a baseline setting in multi-agent reinforcement learning (MARL) with continuous spaces. One feature of our algorithms is that during the learning phase, a certain level of robustness/risk-sensitivity of the controller is preserved, which we termed as the implicit regularization property, and is an essential requirement in safety-critical control systems.

Modeling the arrival process to an Emergency Department (ED) is the first step of all studies dealing with the patient flow within the ED. Many of them focus on the increasing phenomenon of ED overcrowding, which is afflicting hospitals all over the world. Since Discrete Event Simulation models are often adopted with the aim to assess solutions for reducing the impact of this problem, proper nonstationary processes are taken into account to reproduce time-dependent arrivals. Accordingly, an accurate estimation of the unknown arrival rate is required to guarantee the reliability of the results. In this work, an integer nonlinear black-box optimization problem is solved to determine the best piecewise constant approximation of the time-varying arrival rate function, by finding the optimal partition of the 24 hours into a suitable number of non equally spaced intervals. The black-box constraints of the optimization problem make the feasible solutions satisfy proper statistical hypotheses; these ensure the validity of the nonhomogeneous Poisson assumption about the arrival process, commonly adopted in the literature, and prevent mixing overdispersed data for model estimation. The cost function includes a fit error term for the solution accuracy and a penalty term to select an adequate degree of regularity of the optimal solution. To show the effectiveness of this methodology, real data from one of the largest Italian hospital EDs are used.

In this paper, we present difference of convex algorithms for solving bilevel programs in which the upper level objective functions are difference of convex functions, and the lower level programs are fully convex. This nontrivial class of bilevel programs provides a powerful modelling framework for dealing with applications arising from hyperparameter selection in machine learning. Thanks to the full convexity of the lower level program, the value function of the lower level program turns out to be convex and hence the bilevel program can be reformulated as a difference of convex bilevel program. We propose two algorithms for solving the reformulated difference of convex program and show their convergence under very mild assumptions. Finally we conduct numerical experiments to a bilevel model of support vector machine classification.

Does a new product spread faster among heterogeneous or homogeneous consumers? We analyze this question using the stochastic discrete Bass model, in which consumers may differ in their individual external influence rates { p j } and in their individual internal influence rates { q j } . When the network is complete and the heterogeneity is only manifested in { p j } or only in { q j } , it always slows down the diffusion, compared to the corresponding homogeneous network. When, however, consumers are heterogeneous in both { p j } and { q j } , heterogeneity slows down the diffusion in some cases, but accelerates it in others. Moreover, the dominance between the heterogeneous and homogeneous adoption levels is global in time in some cases, but changes with time in others. Perhaps surprisingly, global dominance between two networks is not always preserved under "additive transformations", such as adding an identical node to both networks. When the network is not complete, the effect of heterogeneity depends also on its spatial distribution within the network.