1: \begin{abstract}

2:     We study sampling algorithms for $\beta$-ensembles with time complexity less than cubic in the cardinality of the ensemble.

3:     Following \cite{DuEd02}, we see the ensemble as the eigenvalues of a random tridiagonal matrix, namely a random Jacobi matrix.

4:     First, we provide a unifying and elementary treatment of the tridiagonal models associated to the three classical Hermite, Laguerre and Jacobi ensembles.

5:     For this purpose, we use simple changes of variables between successive reparametrizations of the coefficients defining the tridiagonal matrix.

6:     Second, we derive an approximate sampler for the simulation of $\beta$-ensembles, and illustrate how fast it can be for polynomial potentials.

7:     This method combines a Gibbs sampler on Jacobi matrices and the diagonalization of these matrices.

8:     In practice, even for large ensembles, only a few Gibbs passes suffice for the marginal distribution of the eigenvalues to fit the expected theoretical distribution.

9:     When the conditionals in the Gibbs sampler can be simulated exactly, the same fast empirical convergence is observed for the fluctuations of the largest eigenvalue.

10:     Our experimental results support a conjecture by \citet[][]{KrRiVi16}, that the Gibbs chain on Jacobi matrices of size $N$ mixes in $\mathcal{O}(\log N)$.

11: \end{abstract}

12: