کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
271324 | 504992 | 2012 | 4 صفحه PDF | دانلود رایگان |

We present an integrated framework for (real-time) pattern recognition in fusion data. The main premise is the inherent probabilistic nature of measurements of plasma quantities. We propose the geodesic distance on probabilistic manifolds as a similarity measure between data points. Substructure induced by data dependencies may further reduce the dimensionality and redundancy of the data set. We present an application to confinement mode classification, showing the distinct advantage obtained by considering the measurement uncertainty and its geometry.
► We present an integrated framework for pattern recognition in fusion data.
► We model measurement uncertainty through an appropriate probability distribution.
► We use the geodesic distance on probabilistic manifolds as a similarity measure.
► We apply the framework to confinement mode classification.
► The classification accuracy benefits from uncertainty information and its geometry.
Journal: Fusion Engineering and Design - Volume 87, Issue 12, December 2012, Pages 2068–2071