1: \begin{abstract}

2: In this paper we study the frequentist convergence rate for the Latent Dirichlet Allocation \citep{blei2003latent} topic models.

3: We show that the maximum likelihood estimator converges to one of the finitely many equivalent parameters in Wasserstein's distance metric

4: at a rate of $n^{-1/4}$ without assuming separability or non-degeneracy of the underlying topics and/or the existence of more than three words per document,

5: thus generalizing the previous works of \cite{anandkumar2012spectral,anandkumar2014tensor} from an information-theoretical perspective.

6: We also show that the $n^{-1/4}$ convergence rate is optimal in the worst case.

7: 

8: \textbf{Keywords}: Latent Dirichlet Allocation, topic models, maximum likelihood, rates of convergence

9: \end{abstract}

10: