Modification of Semi-supervised Algorithm Based on Gaussian Random Fields and Harmonic Functions

In this paper we propose an improvement for a semi-supervised learning algorithm based on Gaussian random fields and harmonic functions. Semi-supervised functions is graph-based method that uses data point similarity to connect unlabeled points with labeled points, thus achieving label propagation. The proposed concerns the way of determining between two by using hybrid RBF-kNN kernel. This makes more resilient noise propagation locality-aware. was tested five synthetic datasets. Results indicate there no datasets big margin classes, however in low approach kernel outperforms existing algorithms simple