Learn More
In this paper, we describe how we address the ICML 2004 Physiological Data Modeling Contest. For the gender prediction task, we employ 5 off-the-shelf machine learning methods: decision tree, neural networks, naive bayes, logistic regression, and Support Vector Machines. We use neural networks for the context prediction tasks. Most of the methods perform(More)
According to mathematical properties of continuous function, the universal approximation properties of fuzzy mapping in only one interval are analyzed, and some related theorems are proposed and proved. Then, the sufficient conditions of the general T-S fuzzy systems with universal approximation are derived from these theorems, which helps establish the new(More)
Project Title: Support Vector Machines for Multiple Instance Learning PI: T. Hofmann Participants: Stuart Andrews and Thomas Hofmann Abstract: Multiple Instance Learning (MIL) is an important generalization of standard supervised binary classification. In MIL labels are not available for individual training patterns, but are associated with sets of(More)
A way to deal with the rule-explosion problem is to use the hierarchical fuzzy systems. In this paper, the hierarchical fuzzy system is constructed first and the property of the hierarchical system is discussed. Then the error properties of the polynomial and hierarchical polynomial are given. Based on these properties, the universal approximation property(More)
Face recognition is a typical problem of pattern recognition and machine learning. Among these approaches to the problem of face recognition, subspace analysis gives the most promising results, and becomes one of the most popular methods. This paper researches typical subspace analysis approaches, based on the introduction of main approaches of linear(More)