Gradient boosting machines are a family of powerful machine-learning techniques that have shown considerable success in a wide range of practical applications. They are highly customizable to the particular needs of the application, like being learned with respect to different loss functions. This article gives a tutorial introduction into the methodology of gradient boosting methods with a strong focus on machine learning aspects of modeling. A theoretical information is complemented with descriptive examples and illustrations which cover all the stages of the gradient boosting model design. Considerations on handling the model complexity are discussed. Three practical examples of gradient boosting applications are presented and comprehensively analyzed.
The problem of sensor fault detection is important part of the overall system health estimation. Without assuring correct sensor readings it is impossible to make any conclusion about system status. On the other hand, it is essential not to confuse sensor faults and abnormal sensor behavior caused by other reasons, such as dynamically changing system condition or fault in the system. In the current paper the problem of distinguishing between sensor faults and system dynamics is investigated. This problem cannot be solved by validating each sensor signal independently; it requires joint analysis of all sensors. The particular case of several redundant of wellcorrelated sensors is considered. The proposed solution consists of two steps. First, multiresolution analysis, a powerful waveletbased signal processing technique, is applied to detect signal changes both at high and low frequency scales. Then, found changes are inspected on different wavelet detail levels and decision is made whether these changes are true sensor faults or are caused by system dynamics. The proposed method is tested on data from gas turbine power plant.
Fuzzy neural networks are a powerful machine learning technique, that can be used in a large number of applications. Proper learning of fuzzy neural networks requires a lot of computational effort and the fuzzy-rule designs of these networks suffer from the curse of dimensionality. To alleviate these problems, a simplified fuzzy neural network is presented. The proposed simplified network model can be efficiently initialized with considerably high predictive power. We propose the ensembling approach, thus, using the new simplified neural network models as the type of a general-purpose fuzzy base-learner. The new base-learner properties are analyzed and the practical results of the new algorithm are presented on the robotic hand controller application.
Abstract-The problem of optimal fusion of several predictive machine learning regression models is considered. The method of combining different predictive models based on additive fuzzy systems is presented. The framework of model fusion based on fuzzy neural networks is described and the appropriate algorithms are derived. Learning process justifications and the requirement of the separate fusion set are discussed. The presented models are supported with the real world application example of robotic hand control.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.