Search results for "Function approximation"

Moving Least Squares Innovative Strategies For Sheet Forming Design

2011

In the last years a great interest in optimization algorithms aimed to design forming processes was demonstrated by many researches. Proper design methodologies to reduce times and costs have to be developed mostly based on computer aided procedures. Response surface methods (RSM) proved their effectiveness in the recent years also for the application in sheet metal forming aiming to reduce the number of numerical simulations. Actually, the main drawback of such method is the number of direct problem to be solved in order to reach good function approximations. A very interesting aspect in RSM application regards the possibility to build response surfaces basing on moving least squares appro…

A contribution on the optimization strategies based on moving least squares approximation for sheet metal forming design

2012

Computer-aided procedures to design and optimize forming processes are, nowadays, crucial research topics since industrial interest in costs and times reduction is always increasing. Many researchers have faced this research challenge with various approaches. Response surface methods (RSM) are probably the most known approaches since they proved their effectiveness in the recent years. With a peculiar attention to sheet metal forming process design, RSM should offer the possibility to reduce the number of numerical simulations which in many cases means to reduce design times and complexity. Actually, the number of direct problems (FEM simulations) to be solved in order to reach good functio…

A Comparison between Three Meta-Modeling Optimization Approaches to Design a Tube Hydroforming Process

2012

Computer aided procedures to design and optimize forming processes have become crucial research topics as the industrial interest in cost and time reduction has been increasing. A standalone numerical simulation approach could make the design too time consuming while meta-modeling techniques enables faster approximation of the investigated phenomena, reducing the simulation time. Many researchers are, nowadays, facing such research challenge by using various approaches. Response surface method (RSM) is probably the most known one, since its effectiveness was demonstrated in the past years. The effectiveness of RSM depends both on the definition of the Design of Experiments (DoE) and the acc…

Exploiting Numerical Behaviors in SPH.

2010

Smoothed Particle Hydrodynamics is a meshless particle method able to evaluate unknown field functions and relative differential operators. This evaluation is done by performing an integral representation based on a suitable smoothing kernel function which, in the discrete formulation, involves a set of particles scattered in the problem domain. Two fundamental aspects strongly characterizing the development of the method are the smoothing kernel function and the particle distribution. Their choice could lead to the so-called particle inconsistency problem causing a loose of accuracy in the approximation; several corrective strategies can be adopted to overcome this problem. This paper focu…

Consistency Restoring in SPH for Trigonometric Functions Approximation

2009

Experiments in Value Function Approximation with Sparse Support Vector Regression

2004

We present first experiments using Support Vector Regression as function approximator for an on-line, sarsa-like reinforcement learner. To overcome the batch nature of SVR two ideas are employed. The first is sparse greedy approximation: the data is projected onto the subspace spanned by only a small subset of the original data (in feature space). This subset can be built up in an on-line fashion. Second, we use the sparsified data to solve a reduced quadratic problem, where the number of variables is independent of the total number of training samples seen. The feasability of this approach is demonstrated on two common toy-problems.

Multi-dimensional Function Approximation and Regression Estimation

2002

In this communication, we generalize the Support Vector Machines (SVM) for regression estimation and function approximation to multi-dimensional problems. We propose a multi-dimensional Support Vector Regressor (MSVR) that uses a cost function with a hyperspherical insensitive zone, capable of obtaining better predictions than using an SVM independently for each dimension. The resolution of the MSVR is achieved by an iterative procedure over the Karush-Kuhn-Tucker conditions. The proposed algorithm is illustrated by computers experiments.

A Support Vector Machine Signal Estimation Framework

2018

Support vector machine (SVM) were originally conceived as efficient methods for pattern recognition and classification, and the SVR was subsequently proposed as the SVM implementation for regression and function approximation. Nowadays, the SVR and other kernel‐based regression methods have become a mature and recognized tool in digital signal processing (DSP). This chapter starts to pave the way to treat all the problems within the field of kernel machines, and presents the fundamentals for a simple, framework for tackling estimation problems in DSP using support vector machine SVM. It outlines the particular models and approximations defined within the framework. The chapter concludes wit…

An adaption mechanism for the error threshold of XCSF

2020

Learning Classifier System (LCS) is a class of rule-based learning algorithms, which combine reinforcement learning (RL) and genetic algorithm (GA) techniques to evolve a population of classifiers. The most prominent example is XCS, for which many variants have been proposed in the past, including XCSF for function approximation. Although XCSF is a promising candidate for supporting autonomy in computing systems, it still must undergo parameter optimization prior to deployment. However, in case the later deployment environment is unknown, a-priori parameter optimization is not possible, raising the need for XCSF to automatically determine suitable parameter values at run-time. One of the mo…

Merging Features from Green's Functions and Time Dependent Density Functional Theory: A Route to the Description of Correlated Materials out of Equil…

2016

We propose a description of nonequilibrium systems via a simple protocol that combines exchange-correlation potentials from density functional theory with self-energies of many-body perturbation theory. The approach, aimed to avoid double counting of interactions, is tested against exact results in Hubbard-type systems, with respect to interaction strength, perturbation speed and inhomogeneity, and system dimensionality and size. In many regimes, we find significant improvement over adiabatic time dependent density functional theory or second Born nonequilibrium Green's function approximations. We briefly discuss the reasons for the residual discrepancies, and directions for future work.