Description:Conformal Prediction Methods
Abstract:Conformal Prediction Methods, Development of novel pattern recognition methods based on the algorithmic theory of randomness that, unlike traditional machine learning algorithms, accompany their predictions with confidence measures. Currently this work is focused on developing the Neural Network and Genetic Algorithm conformal predictors. The Grid is very useful in the experimental stages of this work, which involve application of the developed methods on benchmark datasets and evaluation of their results.
This work is part of a project, which is funded by the Cyprus Research Promotion Foundation, titled ”Development of New Conformal Prediction Methods with Applications in Medical Diagnosis”. As its name suggests, a later part of this project consists of the application of the developed methods on medical data. More specifically, the problem under investigation is the diagnosis of children suffering from acute abdominal pain. The Grid will be even more useful in this part of the project, which requires a large number of experiments.