1: \begin{abstract}

2: Nonnegative matrix factorization (NMF) is a standard linear dimensionality reduction technique for nonnegative data sets. 

3: In order to measure the discrepancy between the input data and the low-rank approximation, the Kullback-Leibler (KL) divergence  is one of the most widely used objective function for NMF. 

4: It corresponds to the maximum likehood estimator when the underlying statistics of the observed data sample follows a Poisson distribution, and KL NMF is particularly meaningful for count data sets, such as documents or images. 

5: %In this paper, we consider NMF where the objective function minimizes the  between the input data and the low-rank approximation. 

6: In this paper, %we focus on KL NMF, and our contribution is threefold. 

7: %First, 

8: we first collect important properties of the KL objective function that are essential to study the convergence of KL NMF algorithms. 

9: Second, together with reviewing existing algorithms for solving KL NMF, we propose three new algorithms that guarantee the non-increasingness of the objective function. We also provide a global convergence guarantee

10: for one of our proposed algorithms. 

11: Finally, we conduct extensive numerical experiments to provide a comprehensive picture of the performances of the KL NMF algorithms.  

12: %on different types of data sets and under differ. 

13: \keywords{nonnegative matrix factorization \and 

14:   Kullback-Leibler divergence \and 

15:   Poisson distribution \and 

16:   algorithms}

17: \end{abstract}

18: