NMF-based Models for Tumor Clustering: A Systematic Comparison

Nonnegative Matrix Factorization (NMF) is one of the famous unsupervised learning models. In this paper, we give a short survey on NMF-related models, including K-means, Probabilistic Latent Semantic Indexing etc. and present a new Posterior Probabilistic Clustering model, and compare their numerical experimental results on five real microarray data. The results show that i) NMF using with K-L divergence objective function has better clustering performance; ii) Our purposed PPC model is among the best.