Predicting the functional consequences of genetic variants is a challenging problem, especially for non-coding variants. Projects such as ENCODE and Roadmap Epigenomics make available various epigenetic features genome-wide in over a hundred tissues and cell types. In addition, recent developments in high-throughput assays to assess the functional impact of variants in regulatory regions can lead to the generation of high quality data on the functional effects of selected variants. We propose here a semi-supervised approach, GenoNet, to jointly utilize experimentally confirmed regulatory variants (labeled variants), millions of unlabeled variants genome-wide, and more than a thousand cell type/tissue specific epigenetic annotations to predict functional consequences of non-coding variants. Through the application to several experimental datasets, we demonstrate that the proposed method significantly improves prediction accuracy compared to existing functional prediction methods.