Relevance of support vector machines for stochastic mechanics

TLDR: The utility of the distinguishing features of the support vector method for two important tasks in stochastic mechanics: the learning from random samples in a Monte Carlo simulation context and the possibility of defining a reliability index that characterizes arbitrary safe domains.

Abstract: Publisher Summary This chapter focuses on the relevance of support vector machines for stochastic mechanics. The last years have witnessed a growing research on applications of artificial intelligence algorithms in several fields of computational mechanics. Most applications concern genetic algorithms, fuzzy set reasoning, and neural networks. Few applications have been reported on the kernel methods, which is a family of artificial learning algorithms that are distinguished by their use of kernels. Kernel methods and support vector machines have emerged as a potent artificial intelligence alternative to neural networks for complicated tasks, especially those concerning image analysis. This chapter presents a paper that examines learning algorithm with respect to their utility in solving stochastic mechanics problems. It describes the main features of support vector machines and their differences in neural networks. It discusses the utility of the distinguishing features of the support vector method for two important tasks in stochastic mechanics: the learning from random samples in a Monte Carlo simulation context and the possibility of defining a reliability index that characterizes arbitrary safe domains.