1: \begin{abstract}

2: We propose a Block Majorization Minimization \ngi{method} with Extrapolation (BMMe) for solving a class of multi-convex optimization problems. The extrapolation parameters of BMMe are updated using a novel adaptive update rule. By \Vali{showing } that block majorization minimization can be reformulated as a block mirror descent method, with the Bregman divergence adaptively \ngi{updated at each iteration}, we establish subsequential convergence for BMMe. 

3: \ngi{We use this method to design \Vali{efficient} algorithms to \Vali{tackle} nonnegative matrix factorization problems with \Vali{the} $\beta$-divergences ($\beta$-NMF) for $\beta\in [1,2]$.  % \hien{actually we don't ``show" (I think ``show"=prove), it is we ``recognize" since we just rewrite BMM in an equivalent form} %These algorithms are multiplicative updates with extrapolation, for which our new results provide convergence guarantees. 

4: %\Vali{These algorithms rely on multiplicative updates along with with extrapolation, for which our new results provide convergence guarantees}. 

5: \Vali{These algorithms, \ngir{which are} multiplicative updates  with extrapolation, benefit from our novel results that offer convergence guarantees.} 

6: \revise{We also empirically \ngir{illustrate} the significant acceleration of BMMe for $\beta$-NMF through extensive experiments}.} 

7: \end{abstract}

8: