TY - GEN
T1 - Block-iterative algorithms for non-negative matrix approximation
AU - Sra, Suvrit
PY - 2008
Y1 - 2008
N2 - In this paper we present new algorithms for non-negative matrix approximation (NMA), commonly known as the NMF problem. Our methods improve upon the well-known methods of Lee & Seung [12] for both the Frobenius norm as well the Kullback-Leibler divergence versions of the problem. For the latter problem, our results are especially interesting because it seems to have witnessed much lesser algorithmic progress as compared to the Frobenius norm NMA problem. Our algorithms are based on a particular block-iterative acceleration technique for EM, which pre- serves the multiplicative nature of the updates and also en- sures monotonicity. Furthermore, our algorithms also nat- urally apply to the Bregman-divergence NMA algorithms of [6]. Experimentally, we show that our algorithms outper- form the traditional Lee/Seung approach most of the time.
AB - In this paper we present new algorithms for non-negative matrix approximation (NMA), commonly known as the NMF problem. Our methods improve upon the well-known methods of Lee & Seung [12] for both the Frobenius norm as well the Kullback-Leibler divergence versions of the problem. For the latter problem, our results are especially interesting because it seems to have witnessed much lesser algorithmic progress as compared to the Frobenius norm NMA problem. Our algorithms are based on a particular block-iterative acceleration technique for EM, which pre- serves the multiplicative nature of the updates and also en- sures monotonicity. Furthermore, our algorithms also nat- urally apply to the Bregman-divergence NMA algorithms of [6]. Experimentally, we show that our algorithms outper- form the traditional Lee/Seung approach most of the time.
UR - http://www.scopus.com/inward/record.url?scp=67049161175&partnerID=8YFLogxK
U2 - 10.1109/ICDM.2008.77
DO - 10.1109/ICDM.2008.77
M3 - Conference contribution
AN - SCOPUS:67049161175
SN - 9780769535029
T3 - Proceedings - IEEE International Conference on Data Mining, ICDM
SP - 1037
EP - 1042
BT - Proceedings - 8th IEEE International Conference on Data Mining, ICDM 2008
T2 - 8th IEEE International Conference on Data Mining, ICDM 2008
Y2 - 15 December 2008 through 19 December 2008
ER -