1: \begin{abstract}

2: \indent Visual rendering of graphs is a key task in the mapping of

3: complex network data. Although most graph drawing algorithms

4: emphasize aesthetic appeal, certain applications such as travel-time

5: maps place more importance on visualization of structural network

6: properties. The present paper advocates two graph embedding approaches

7: with centrality considerations to comply with node hierarchy. The

8: problem is formulated first as one of constrained multi-dimensional

9: scaling (MDS), and it is solved via block coordinate descent

10: iterations with successive approximations and guaranteed convergence

11: to a KKT point. In addition, a regularization term enforcing graph

12: smoothness is incorporated with the goal of reducing edge crossings.

13: A second approach leverages the locally-linear embedding (LLE) 

14: algorithm which assumes that the graph encodes data sampled from

15: a low-dimensional manifold. Closed-form solutions to the resulting

16: centrality-constrained optimization problems are determined yielding

17: meaningful embeddings. Experimental results demonstrate the efficacy

18: of both approaches, especially for visualizing large networks

19:  on the order of thousands of nodes.

20: \end{abstract}

21: