1: \begin{abstract}

2: The performance of the generalized belief propagation algorithm 

3: to compute the noiseless capacity and mutual \mbox{information} rates of 

4: finite-size two-dimensional and three-dimensional run-length limited 

5: constraints is investigated. In both cases, the problem is reduced to

6: estimating the partition function of graphical models with cycles. The

7: partition function is then estimated using the region-based free energy 

8: approximation technique. For each constraint, a method is 

9: proposed to choose the basic regions and to construct the region graph

10: which provides the graphical framework

11: to run the generalized belief propagation algorithm.

12: Simulation results for the noiseless capacity of different constraints as a 

13: function of the size of the channel are reported.

14: In the cases that tight lower and upper bounds on the Shannon capacity exist,

15: convergence to the Shannon capacity is discussed.

16: For noisy constrained channels, 

17: simulation results are reported for

18: mutual information rates as a function of signal-to-noise ratio.

19: \end{abstract}

20: