Software parallelization of a probabilistic classifier based on Venn Prediction: Application to the TJ-II Thomson Scattering

One of the recurring problems encountered in the development of automatic classification problems is the so-called “curse of dimensionality”. Procedures that are computationally manageable in low dimensional spaces can become unfeasible in spaces of hundreds of dimensions due to the need of long computational times. This paper shows the parallelization of a probabilistic classifier based on Venn Predictors (VP). VP determine a probability interval to qualify how accurate and reliable each individual classification is. The parallelized code has been applied to the classification of the images from the CCD camera of the TJ-II Thomson Scattering. The aver- age probability and probability interval are a very efficient prediction from the prediction perspective.