Tianpei Xie's research focuses on robust machine learning from multiple data sources. He has developed algorithms for robust classification in the presence of noisy or corrupted training data, including GEM-MED which jointly performs classification and anomaly detection. He has also developed methods for multi-view learning on statistical manifolds, including CMV-MED which co-regularizes multiple models using a robust consensus measure based on information divergence between probability density functions. Current work involves predicting node attributes in networks by combining network topology and node distributions. He has published papers in major machine learning conferences and journals and maintains websites with details of his research activities.