Showing papers on "Piecewise linear function published in 2002"

Regression trees with unbiased variable selection and interaction detection

Abstract: We propose an algorithm for regression tree construction called GUIDE. It is specifically designed to eliminate variable selection bias, a problem that can undermine the reliability of inferences from a tree structure. GUIDE controls bias by employing chi-square analysis of residuals and bootstrap calibration of signif- icance probabilities. This approach allows fast computation speed, natural ex- tension to data sets with categorical variables, and direct detection of local two- variable interactions. Previous algorithms are not unbiased and are insensitive to local interactions during split selection. The speed of GUIDE enables two further enhancements—complex modeling at the terminal nodes, such as polynomial or best simple linear models, and bagging. In an experiment with real data sets, the prediction mean square error of the piecewise constant GUIDE model is within ±20% of that of CART r � . Piecewise linear GUIDE models are more accurate; with bagging they can outperform the spline-based MARS r � method.

Piecewise linear skeletonization using principal curves

Abstract: Proposes an algorithm to find piecewise linear skeletons of handwritten characters by using principal curves. The development of the method was inspired by the apparent similarity between the definition of principal curves (smooth curves which pass through the "middle" of a cloud of points) and medial axes (smooth curves that run equidistantly from the contours of a character image). The central fitting-and-smoothing step of the algorithm is an extension of the polygonal line algorithm, which approximates principal curves of data sets by piecewise linear curves. The polygonal line algorithm is extended to find principal graphs and complemented with two steps specific to the task of skeletonization: an initialization method to capture the approximate topology of the character, and a collection of restructuring operations to improve the structural quality of the skeleton produced by the initialization method. An advantage of our approach over existing methods is that we optimize the skeleton graph by minimizing an intuitive and explicit objective function that captures the two competing criteria of smoothing the skeleton and fitting it closely to the pixels of the character image. We tested the algorithm on isolated handwritten digits and images of continuous handwriting. The results indicated that the proposed algorithm can find a smooth medial axis in the great majority of a wide variety of character templates and that it substantially improves the pixel-wise skeleton obtained by traditional thinning methods.

Piecewise Linear Control Systems: A Computational Approach

Abstract: This book presents a computational approach to the analysis of nonlinear and uncertain systems. The main focus is systems with piecewise linear dynamics. The class of piecewise linear systems examined has nonlinear, possibly discontinuous dynamics, and allows switching rules that incorporate memory and logic. These systems may exhibit astonishingly complex behaviors. Some aspects of the successful theory of linear systems and quadratic criteria are extended here to piecewise linear systems and piecewise quadratic criteria. The book also describes numerical procedures for assessing stability, computing induced gains, and solving optimal control problems for piecewise linear systems. These developments enable researchers to analyze a large and practically important class of control systems that are not easily dealt with when using other techniques. (Less)

Optimal control of constrained, piecewise affine systems with bounded disturbances

Abstract: The solution to the problem of optimal control of piecewise affine systems with a bounded disturbance is characterised Results that allow one to compute the value function, its domain (robustly controllable set) and the optimal control law are presented The tools that are employed include dynamic programming, polytopic set algebra and parametric programming When the cost is time (robust time-optimal control problem) or the stage cost is piecewise affine (robust optimal and robust receding horizon control problems), the value function and the optimal control law are both piecewise affine and each robustly controllable set is the union of a finite set of polytopes Conditions on the cost and constraints are also proposed in order to ensure that the optimal control laws are robustly stabilising

Small data oscillation implies the saturation assumption

Abstract: The saturation assumption asserts that the best approximation error in \(H^1_0\) with piecewise quadratic finite elements is strictly smaller than that of piecewise linear finite elements. We establish a link between this assumption and the oscillation of \(f=-\Delta u\), and prove that small oscillation relative to the best error with piecewise linears implies the saturation assumption. We also show that this condition is necessary, and asymptotically valid provided \(f\in L^2\).

On multi-parametric nonlinear programming and explicit nonlinear model predictive control

Abstract: A numerical algorithm for approximate multi-parametric nonlinear programming is developed. It allows approximate solutions to nonlinear optimization problems to be computed as explicit piecewise linear functions of the problem parameters. In control applications such as nonlinear constrained model predictive control this allows efficient online implementation in terms of an explicit piecewise linear state feedback without any real-time optimization.

Generating chaos with a switching piecewise-linear controller.

Abstract: This paper introduces a new chaos generator, a switching piecewise-linear controller, which can create chaos from a three-dimensional linear system within a wide range of parameter values. Basic dynamical behaviors of the chaotic controlled system are investigated in some detail.

A simple algorithm for localized construction of non‐matching structural interfaces

Abstract: A simple and effective algorithm for the modular construction of non-matched interfaces is presented for the partitioned solution of large-scale structural problems. The formulation is based on a recently developed four-field variational principle, which introduces a connection frame between the interfaced partitions. A key result of the present study is a frame nodal placement criterion that uniquely determines the frame discretization into piecewise linear elements so that the interface patch test condition is satisfied a priori. The method is demonstrated with several 2D and 3D example problems.

Modeling and parameter identification of systems with multisegment piecewise-linear characteristics

Abstract: This paper deals with the modeling and parameter identification of nonlinear systems having multi-segment piecewise-linear characteristics. The decomposition of the corresponding mapping provides a new form of multi-segment nonlinearity representation, leading to an output equation where all the parameters to be estimated are separated. Hence, an iterative method with internal variable estimation can be applied for parameter identification using input/output data records. The only required a-priori knowledge of the nonlinear characteristic represents the limits for the domain partition. The proposed model of given static nonlinearity is also incorporated into the Hammerstein model. Examples of parameter identification for static and dynamic systems with multi-segment piecewise-linear characteristics are presented.

Controller design and analysis of uncertain piecewise-linear systems

Abstract: This paper presents controller design and analysis methods for uncertain piecewise-linear systems based on a piecewise-smooth Lyapunov function. The basic idea of the proposed approaches is to construct controllers for the uncertain piecewise-linear systems in such a way that a piecewise-continuous Lyapunov function can he used to establish the global stability or the global stability with H/sub /spl infin// performance of the resulting closed loop control systems. It is shown that the control laws can be obtained by solving a set of linear matrix inequalities that are numerically feasible with commercially available software. An example is given to illustrate the application of the proposed methods.

On simplifying and classifying piecewise-linear systems

Abstract: A basic methodology to understand the dynamical behavior of a system relies on its decomposition into simple enough functional blocks. In this work, following that idea, we consider a family of piecewise-linear systems that can be written as a feedback structure. By using some results related to control systems theory, a simplifying procedure is given. In particular, we pay attention to obtain equivalent state equations containing both a minimum number of nonzero coefficients and a minimum number of nonlinear dynamical equations (canonical forms). Two new canonical forms are obtained, allowing to classify the members of the family in different classes. Some consequences derived from the above simplified equations are given. The state equations of different electronic oscillators with two or three state variables and two or three linear regions are studied, illustrating the proposed methodology.

Max-Min Representation of Piecewise Linear Functions

Abstract: It is shown that a piecewise linear function on a convex domain in R d can be represented as a boolean polynomial in terms of its linear components.

Approximation Classes for Adaptive Methods

Abstract: Adaptive Finite Element Methods (AFEM) are numerical proce- dures that approximate the solution to a partial differential equation (PDE) by piecewise polynomials on adaptively generated triangulations. Only re- cently has any analysis of the convergence of these methods (10, 13) or their rates of convergence (2) become available. In the latter paper it is shown that a certain AFEM for solving Laplace's equation on a polygonal domain ⊂ R 2 based on newest vertex bisection has an optimal rate of convergence in the following sense. If, for some s > 0 and for each n = 1,2, . . ., the solu- tion u can be approximated in the energy norm to order O(n s ) by piecewise linear functions on a partition P obtained from n newest vertex bisections, then the adaptively generated solution will also use O(n) subdivisions (and floating point computations) and have the same rate of convergence. The question arises whether the class of functions A s with this approximation rate can be described by classical measures of smoothness. The purpose of the present paper is to describe such approximation classes A s by Besov smoothness.

Nonparametric Estimation with Nonlinear Budget Sets

Abstract: Choice models with nonlinear budget sets provide a precise way of accounting for the nonlinear tax structures present in many applications. In this paper we propose a nonparametric approach to estimation of these models. The basic idea is to think of the choice, in our case hours of labor supply, as being a function of the entire budget set. Then we can do nonparametric regression where the variable in the regression is the budget set. We reduce the dimensionality of this problem by exploiting structure implied by utility maximization with piecewise linear convex budget sets. This structure leads to estimators where the number of segments can differ across observations and does not affect accuracy. We give consistency and asymptotic normality results for these estimators. The usefulness of the estimator is demonstrated in an empirical example, where we find it has a large impact on estimated effects of the Swedish tax reform.

On the dynamic analysis of piecewise-linear networks

Abstract: Piecewise-linear (PL) modeling is often used to approximate the behavior of nonlinear circuits. One of the possible PL modeling methodologies is based on the linear complementarity problem, and this approach has already been used extensively in the circuits and systems community for static networks. In this paper, the object of study is dynamic electrical circuits that can be recast as linear complementarity systems, i.e. as interconnections of linear time-invariant differential equations and complementarity conditions (ideal diode characteristics). A mathematically precise framework is developed that formalizes the mixed discrete and continuous behavior of these switched networks. Within this framework, the fundamental question of well-posedness (existence and uniqueness of solution trajectories given an initial condition) is studied and additional properties of the behavior are derived. For instance, a full characterization is presented of the inconsistent states.

Observer design for a class of piece-wise affine systems

Abstract: In this paper we propose an observer design procedure for a class of bi-modal piece-wise affine systems. The designed observers have the characteristic feature that they do not require information on the currently active dynamics of the piecewise linear system. A design procedure which guarantees global asymptotic stability of the estimation error is presented. It is shown that the applicability of the presented procedure is limited to continuous piece-wise affine systems. Therefore, we present an observer design procedure, applicable also to discontinuous systems, which guarantees that the estimation error is bounded, with respect to the state bounds, asymptotically. Sliding motions in the observed system, and the observer are discussed. The presented theory is illustrated with an example.

Finite Volume Methods for Elliptic PDE's: A New Approach

Abstract: We consider a new formulation for finite volume element methods, which is satisfied by known finite volume methods and it can be used to introduce new ones. This framework results by approximating the test function in the formulation of finite element method. We analyze piecewise linear conforming or nonconforming approximations on nonuniform triangulations and prove optimal order H 1-norm and L 2-norm error estimates.

Designing 2D Vector Fields of Arbitrary Topology

Abstract: We introduce a scheme of control polygons to design topological skeletons for vector fields of arbitrary topology. Based on this we construct piecewise linear vector fields of exactly the topology specified by the control polygons. This way a controlled construction of vector fields of any topology is possible. Finally we apply this method for topology-preserving compression of vector fields consisting of a simple topology.

Parametric query optimization for linear and piecewise linear cost functions

Abstract: The cost of a query plan depends on many parameters, such as predicate selectivities and available memory, whose values may not be known at optimization time. Parametric query optimization (PQO) optimizes a query into a number of candidate plans, each optimal for some region of the parameter space. We first propose a solution for the PQO problem for the case when the cost functions are linear in the given parameters. This solution is minimally intrusive in the sense that an existing query optimizer can be used with minor modifications: the solution invokes the conventional query optimizer multiple times, with different parameter values. We then propose a solution for the PQO problem for the case when the cost functions are piecewise-linear in the given parameters. The solution is based on modification of an existing query optimizer. This solution is quite general, since arbitrary cost functions can be approximated to piecewise linear form. Both the solutions work for an arbitrary number of parameters.

Mixed finite element methods for unilateral problems: convergence analysis and numerical studies

Abstract: In this paper, we propose and study different mixed variational methods in order to approximate with finite elements the unilateral problems arising in contact mechanics. The discretized unilateral conditions at the candidate contact interface are expressed by using either continuous piecewise linear or piecewise constant Lagrange multipliers in the saddle-point formulation. A priori error estimates are established and several numerical studies corresponding to the different choices of the discretized unilateral conditions are achieved.

Conditioning of convex piecewise linear stochastic programs

Abstract: In this paper we consider stochastic programming problems where the objective function is given as an expected value of a convex piecewise linear random function. With an optimal solution of such a problem we associate a condition number which characterizes well or ill conditioning of the problem. Using theory of Large Deviations we show that the sample size needed to calculate the optimal solution of such problem with a given probability is approximately proportional to the condition number.

Design of stabilizing switching control laws for discrete- and continuous-time linear systems using piecewise-linear Lyapunov functions

Abstract: In this paper, the stability of switched linear systems is investigated using piecewise linear Lyapunov functions. In particular, we identify classes of switching sequences that result in stable trajectories. Given a switched linear system, we present a systematic methodology for computing switching laws that guarantee stability based on the matrices of the system. In the proposed approach, we assume that each individual subsystem is stable and admits a piecewise linear Lyapunov function. Based on these Lyapunov functions, we compose 'global' Lyapunov functions that guarantee stability of the switched linear system. A large class of stabilizing switching sequences for switched linear systems is characterized by computing conic partitions of the state space. The approach is applied to both discrete-time and continuous-time switched linear systems.

A posteriori error estimation for reduced-basis approximation of parametrized elliptic coercive partial differential equations : “convex inverse” bound conditioners

Abstract: We present a technique for the rapid and reliable prediction of linear-functional outputs of elliptic coercive partial differential equations with affine parameter dependence. The essential components are (i ) (provably) rapidly convergent global reduced-basis approximations – Galerkin projection onto a space W N spanned by solutions of the governing partial differential equation at N selected points in parameter space; (ii ) a posteriori error estimation – relaxations of the error-residual equation that provide inexpensive bounds for the error in the outputs of interest; and ( iii ) off-line/on-line computational procedures – methods which decouple the generation and projection stages of the approximation process. The operation count for the on-line stage – in which, given a new parameter value, we calculate the output of interest and associated error bound – depends only on N (typically very small) and the parametric complexity of the problem; the method is thus ideally suited for the repeated and rapid evaluations required in the context of parameter estimation, design, optimization, and real-time control. In our earlier work we develop a rigorous a posteriori error bound framework for reduced-basis approximations of elliptic coercive equations. The resulting error estimates are, in some cases, quite sharp: the ratio of the estimated error in the output to the true error in the output, or effectivity , is close to (but always greater than) unity. However, in other cases, the necessary “bound conditioners” – in essence, operator preconditioners that (i ) satisfy an additional spectral “bound” requirement, and (ii ) admit the reduced-basis off-line/on-line computational stratagem – either can not be found, or yield unacceptably large effectivities. In this paper we introduce a new class of improved bound conditioners: the critical innovation is the direct approximation of the parametric dependence of the inverse of the operator (rather than the operator itself); we thereby accommodate higher-order (e.g. , piecewise linear) effectivity constructions while simultaneously preserving on-line efficiency. Simple convex analysis and elementary approximation theory suffice to prove the necessary bounding and convergence properties.

An analytical shock-fitting algorithm for LWR kinematic wave model embedded with linear speed–density relationship

Abstract: In this paper, we propose a simple and efficient shock-fitting solution algorithm for the LWR model assuming a linear speed–density relationship or parabolic fundamental diagram. The solution is exact if the boundary conditions for density variable on the spatial axis are piecewise linear and those on the time axis are piecewise constant. Discontinuities are explicitly handled. The method utilizes the concept that for a linear speed–density relationship, a linear density variation along the spatial axis remains linear if not interrupted by shocks. Explicit expressions for the nonlinear shock path trajectory between two linear density functions are also derived. Two numerical examples are used to illustrate the effectiveness of the proposed method.