Hierarchical Data Representation Model - Multi-layer NMF

In this paper, we propose a data representation model that demonstrates hierarchical feature learning using nsNMF. We extend unit algorithm into several layers to take step-by-step approach in learning. Experiments with document and image data successfully demonstrated feature hierarchies. In addition, we showed that taking hierarchical steps also guarantees learning of more meaningful features which leads to better distributed representations, and improved performance in classification and reconstruction tasks, which supports benefits of further application of our proposed data representation model.