Statistical learning theory

About: Statistical learning theory is a research topic. Over the lifetime, 1618 publications have been published within this topic receiving 158033 citations.

Path classification by stochastic linear recurrent neural networks.

Abstract: We investigate the functioning of a classifying biological neural network from the perspective of statistical learning theory, modelled, in a simplified setting, as a continuous-time stochastic recurrent neural network (RNN) with identity activation function. In the purely stochastic (robust) regime, we give a generalisation error bound that holds with high probability, thus showing that the empirical risk minimiser is the best-in-class hypothesis. We show that RNNs retain a partial signature of the paths they are fed as the unique information exploited for training and classification tasks. We argue that these RNNs are easy to train and robust and back these observations with numerical experiments on both synthetic and real data. We also exhibit a trade-off phenomenon between accuracy and robustness.

Application of support vector machine in industrial process

Abstract: Support vector machine(SVM) is new learning machine based on statistical learning theory, which is a kind of learning algorithm focused on small sample It has improved the ability of generation greatly and solved the over-fitting problem of neural network successfully by using the principle of structural risk minimization Recently SVM has been used in the field of pattern recognition, however much less in the field of industrial process This article firstly introduces the development of SVM from the following sides : theory research, algorithm configuration, parameter selection and expanding SVM, then analyzes the applications of SVM to industrial process, such as fault diagnosis, process modeling, system identification, and nonlinear control and so on Finally, the future research directions are pointed out

Maximum Entropy Approach for Linear SVM Optimization Problems

Abstract: Support vector machine(SVM) is a new class of machine learning algorithms,which has been applied to many real-world problems,such as pattern classification, regression analysis and density estimation.This paper presents a new approach to solve linear support vector machines optimization problem.Based on statistical learning theory and optimization theory,unconstrained optimization models for support vector machines are built,and an approximate algorithm-maximum entropy method is given.Primary numerical results illustrate that maximum entropy method for support vector machines is feasible and effective.

Study on Support Vector Machines Algorithms and Its Application to Power Systems

Abstract: The basic statistical learning theory (SLT) and its corresponding algorithms, support vector machines (SVMs), are surveyed, and especially, its latest research results are summarized and studied. By deeply analyzing seven main multi-classification training algorithms of SVMs, including the one-against-rest algorithm, one-against-one algorithm, hierarchy classification, k-class classification, QP-MC-SV algorithm, DDAGSVM algorithm, and sphere structure classification, their respective merits and demerits are found out and listed. Finally, this paper concludes many applications of SVM in the area of fault identification and power systems, specially in the area of power systems transient stability analysis (TSA) and evaluation, fault diagnostics of an electrical machine, accurate fault location in the power transmission line, nonlinear model building of fielding-winding doubly salient generator, flame monitoring, and power system short-term load forecasting and so on. Results demonstrate that SVMs overcome inherent shortcomings of traditional neural network, such as not global optimization, convergence not easily controlling , and difficult network structure designing.