General Piecewise Linear Function Research Articles

Global stability of multi-agent systems with heterogeneous transmission and perception functions

In this brief, we consider an extended opinion dynamic framework by including both transmitting and perceiving behaviors in the agents’ interactions: the first represents how the agent transmits his own opinion to the neighbors, while the latter models how personal features of the agent affect the final perception of external opinions. The agents’ interactions are modeled by general piecewise linear functions that can be heterogeneous and not necessarily monotone, thus generalizing the analytical framework usually considered in the literature, in particular, the well-known interval consensus (Fontanet al., 2020). In the considered novel multi-agent scenario, we formulate sufficient operative LMI conditions to assess global network asymptotic stability for either a consensus or a cluster equilibrium, without necessarily requiring strong network connectivity. The proposed approach is validated through illustrative examples.

Analysis of limited-memory BFGS on a class of nonsmooth convex functions

Abstract The limited-memory BFGS (Broyden-Fletcher-Goldfarb-Shanno) method is widely used for large-scale unconstrained optimization, but its behavior on nonsmooth problems has received little attention. L-BFGS (limited memory BFGS) can be used with or without ‘scaling’; the use of scaling is normally recommended. A simple special case, when just one BFGS update is stored and used at every iteration, is sometimes also known as memoryless BFGS. We analyze memoryless BFGS with scaling, using any Armijo–Wolfe line search, on the function $f(x) = a|x^{(1)}| + \sum _{i=2}^{n} x^{(i)}$, initiated at any point $x_0$ with $x_0^{(1)}\not = 0$. We show that if $a\ge 2\sqrt{n-1}$, the absolute value of the normalized search direction generated by this method converges to a constant vector, and if, in addition, $a$ is larger than a quantity that depends on the Armijo parameter, then the iterates converge to a nonoptimal point $\bar x$ with $\bar x^{(1)}=0$, although $f$ is unbounded below. As we showed in previous work, the gradient method with any Armijo–Wolfe line search also fails on the same function if $a\geq \sqrt{n-1}$ and $a$ is larger than another quantity depending on the Armijo parameter, but scaled memoryless BFGS fails under a weaker condition relating $a$ to the Armijo parameter than that implying failure of the gradient method. Furthermore, in sharp contrast to the gradient method, if a specific standard Armijo–Wolfe bracketing line search is used, scaled memoryless BFGS fails when $a\ge 2 \sqrt{n-1}$regardless of the Armijo parameter. Finally, numerical experiments indicate that the results may extend to scaled L-BFGS with any fixed number of updates $m$, and to more general piecewise linear functions.

A single allocation p-hub median problem with general piecewise-linear costs in arcs

Hubs are used in postal, air passenger transportation, and less-than-truckload (LTL) transportation industries to consolidate, sort, and classify the flows from multiple origins to multiple destinations (OD pairs). Hub-and-spoke networks are commonly designed to exploit economies of scale, and multiple cost functions have been proposed. Hub Location Problems have been used extensively to design hub-and-spoke networks.Notwithstanding the previous efforts in developing accurate cost functions in hub location problems, there is no as today a model flexible enough to represent various practical cost curves. To address this gap, we propose a single-allocation p-hub location problem, in which the cost of the flow on any arc of the network is modeled as a general piecewise-linear function. Inspired by the parcel delivery industry, we allow the existence of routes containing more than three arcs, i.e., any single route may use more than two hubs.The resulting model is hard to solve, even for small instances. We propose a math-heuristic (i.e., a heuristic with exact mathematical programming components), which allows us to solve instances with up to 200 nodes. We obtain solutions considering different cost functions, showing the stability and scalability of the model. Our experiments reveal the dependence between the cost function considered and both hub location and node allocation.

A novel hybrid column generation-metaheuristic approach for the vehicle routing problem with general soft time window

The vehicle routing problem with general soft time window involves designing a set of routes for a fleet of vehicles based at a central depot that is required to service a number of geographically dispersed customers while minimizing the total travel distance and delivery time costs. Delivery time cost function is a general piecewise linear function. In this study, we propose a mathematical model of this problem. Then an efficient hybrid column generation-metaheuristic approach is developed. In the proposed algorithm, the hybridization of column generation (CG) and the metaheuristic is performed in both integrative and collaborative modes. In the integrative phase, a quantum-inspired evolutionary algorithm is used to solve the sub-problems of column generation. In the collaborative phase, the column generation and electromagnetism algorithms are parallelized, and the information from these two algorithms is exchanged to find better solutions. Finally, the performance of the proposed approach is evaluated using a set of modified classic benchmark instances adopted from the literature.

Synchronized production–distribution planning in a single-plant multi-destination network

This paper examines the flow synchronization problem between a manufacturing location and multiple destinations. Multiple products can be shipped from the manufacturing location to different locations via multiple transportation modes. These transportation modes may have different transportation lead times. The transportation costs structure of the different transportation modes offer economies of scale and can be represented by general piecewise linear functions. The production system at the manufacturing location is a serial process with a bottleneck stage. At the bottleneck stage, a predetermined production sequence must be maintained as is the case in some process-based industries. We propose a tight mixed integer programming model for integrated planning of production and distribution in the network. We show that by adding simple valid inequalities and special 0-1 variables, major computational improvements can be achieved when solving this problem with commercial solvers such as Cplex. We also propose a sequential solution approach, based on the independent, but synchronized, solutions of the production and distribution sub-problems. Finally, the solution methods proposed are tested experimentally for realistic problems and the advantage of integrated planning over independent but synchronized planning is assessed.

Multi-item dynamic production-distribution planning in process industries with divergent finishing stages

This paper examines a multi-item dynamic production-distribution planning problem between a manufacturing location and a distribution center. Transportation costs between the manufacturing location and the distribution center offer economies of scale and can be represented by general piecewise linear functions. The production system at the manufacturing location is a serial process with a multiple parallel machines bottleneck stage and divergent finishing stages. A predetermined production sequence must be maintained on the bottleneck machines. A tight mixed-integer programming model of the production process is proposed, as well as three different formulations to represent general piecewise linear functions. These formulations are then used to develop three equivalent mathematical programming models of the manufacturer-distributor flow planning problem. Valid inequalities to strengthen these formulations are proposed and the strategy of adding extra 0–1 variables to improve the branching process is examined. Tests are performed to compare the computational efficiency of these models. Finally, it is shown that by adding valid inequalities and extra 0–1 variables, major computational improvements can be achieved.

Local search algorithms for the rectangle packing problem with general spatial costs

We propose local search algorithms for the rectangle packing problem to minimize a general spatial cost associated with the locations of rectangles. The problem is to pack given rectangles without overlap in the plane so that the maximum cost of the rectangles is minimized. Each rectangle has a set of modes, where each mode specifies the width and height of the rectangle and its spatial cost function. The spatial costs are general piecewise linear functions which can be non-convex and discontinuous. Various types of packing problems and scheduling problems can be formulated in this form. To represent a solution of this problem, a pair of permutations of n rectangles is used to determine their horizontal and vertical partial orders, respectively. We show that, under the constraint specified by such a pair of permutations, optimal locations of the rectangles can be efficiently determined by using dynamic programming. The search for finding good pairs of permutations is conducted by local search and metaheuristic algorithms. We report computational results on various implementations using different neighborhoods, and compare their performance. We also compare our algorithms with other existing heuristic algorithms for the rectangle packing problem and scheduling problem. These computational results exhibit good prospects of the proposed algorithms.

Some extensions of a new method to analyze complete stability of neural networks

In a recent work, a new method has been introduced to analyze complete stability of the standard symmetric cellular neural networks (CNNs), which are characterized by local interconnections and neuron activations modeled by a three-segment piecewise-linear (PWL) function. By complete stability it is meant that each trajectory of the neural network converges toward an equilibrium point. The goal of this paper is to extend that method in order to address complete stability of the much wider class of symmetric neural networks with an additive interconnecting structure where the neuron activations are general PWL functions with an arbitrary number of straight segments. The main result obtained is that complete stability holds for any choice of the parameters within the class of symmetric additive neural networks with PWL neuron activations, i.e., such a class of neural networks enjoys the important property of absolute stability of global pattern formation. It is worth pointing out that complete stability is proved for generic situations where the neural network has finitely many (isolated) equilibrium points, as well as for degenerate situations where there are infinite (nonisolated) equilibrium points. The extension in this paper is of practical importance since it includes neural networks useful to solve significant signal processing tasks (e.g., neural networks with multilevel neuron activations). It is of theoretical interest too, due to the possibility of approximating any continuous function (e.g., a sigmoidal function), using PWL functions. The results in this paper confirm the advantages of the method of Forti and Tesi, with respect to LaSalle approach, to address complete stability of PWL neural networks.