1: \begin{abstract}

2: We propose a fast approximate algorithm for large graph

3: matching. A new projected fixed-point

4: method is defined and a new doubly stochastic projection is adopted to derive the

5: algorithm. Previous graph matching algorithms suffer from high

6: computational complexity and therefore do not have good scalability

7: with respect to graph size. For matching two weighted graphs of $n$

8: nodes, our algorithm has time complexity only $O(n^3)$ per iteration and space

9: complexity $O(n^2)$. In addition to its scalability, our algorithm is

10: easy to implement, robust, and able to match undirected weighted attributed graphs of

11: different sizes. While the convergence rate of previous iterative graph matching algorithms is unknown, our algorithm is theoretically guaranteed to converge at a linear rate. Extensive experiments on large synthetic and real

12: graphs (more than 1,000 nodes) were conducted to evaluate the

13: performance of various algorithms. Results show that in most cases our proposed

14: algorithm achieves better performance than previous state-of-the-art algorithms in terms

15: of both speed and accuracy in large graph matching. In particular, with high

16: accuracy, our algorithm takes only a few seconds (in a PC) to match

17: two graphs of 1,000 nodes.

18: \end{abstract}

19: