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.

Learning theory analysis for association rules and sequential event prediction

Abstract: We present a theoretical analysis for prediction algorithms based on association rules. As part of this analysis, we introduce a problem for which rules are particularly natural, called "sequential...

A Novel Aero-Engine Identification Model Based on Support Vector Machines

Abstract: Because of the strong non-linearity and time-varying properties,and the existence of local minima,overlearning of conventional neural networks, a novel aero-engine identification model is introducedThe model is based on modern statistical learning theory,and Structure Risk Minimization (SRM) principleIt has very good generalization abilityBy solving a quadratic convex programming problem, a global optimum can be automatically foundThe identification model was established based on support vector regression machines using the real flight recorded data as learning samplesThe results show that this method reveals high precision,good robustness and fault-tolerant ability It provides a general way for aero-engine model identification

Structural Risk Minimization Principle Based on Complex Random Samples

Abstract: Statistical Learning Theory is commonly regarded as a sound framework within which we handle a variety of learning problems in presence of small size data samples. It has become a rapidly progressing research area in machine learning. The theory is based on real random samples and as such is not ready to deal with the statistical learning problems involving complex fuzzy random samples, which we may encounter in real world scenarios. This paper explores statistical learning theory based on complex fuzzy random samples. Firstly, the definition of complex fuzzy random variable is introduced. Next the concepts and some properties of the mathematical expectation and independence of complex fuzzy random variables are provided. Secondly, the concepts of annealed entropy, growth function and VC dimension of measurable complex fuzzy set valued functions are proposed, and the bounds on the rate of uniform convergence of learning process based on complex fuzzy random samples are constructed. Thirdly, on the basis of these bounds, the idea of the complex fuzzy structural risk minimization principle is presented. Finally, the consistency of this principle is proven and the bound on the asymptotic rate of convergence is derived.

Statistical Learning Theory Approach for Data Classification with l-diversity

Abstract: Corporations are retaining ever-larger corpuses of personal data; the frequency or breaches and corresponding privacy impact have been rising accordingly. One way to mitigate this risk is through use of anonymized data, limiting the exposure of individual data to only where it is absolutely needed. This would seem particularly appropriate for data mining, where the goal is generalizable knowledge rather than data on specific individuals. In practice, corporate data miners often insist on original data, for fear that they might "miss something" with anonymized or differentially private approaches. This paper provides a theoretical justification for the use of anonymized data. Specifically, we show that a support vector classifier trained on anatomized data satisfying l-diversity should be expected to do as well as on the original data. Anatomy preserves all data values, but introduces uncertainty in the mapping between identifying and sensitive values, thus satisfying l-diversity. The theoretical effectiveness of the proposed approach is validated using several publicly available datasets, showing that we outperform the state of the art for support vector classification using training data protected by k-anonymity, and are comparable to learning on the original data.

Research on Automatic Fingerprint Classification Based on Support Vector Machine

Abstract: Automatic fingerprint classification is an important part of Fingerprint Automatic Identification System (FAIS). Its function is to provide a search system for large size database. Accurate classification can reduce searching time and expediate matching speed. Support Vector Machine (SVM) is a new learning technique based on Statistical Learning Theory (SLT). SVM was originally developed for two-class classification. It was extended to solve multi-class classification problem. A hierarchical SVM with clustering algorithm based on stepwise decomposition was established to intellectively classify 5 classes of fingerprints. The design principle was proposed and the classification algorithm was implemented. SVM not only has more solid theoretical foundation, it also has greater generalization ability as our experiment demonstrates. The experimental results show that SVM is effective and surpasses other classical classification techniques.