1: \begin{abstract}

2: Linear programming has been one of the most extensively studied 

3: branches in mathematics and has been found many applications in

4: science and engineering.

5: Since the beginning of the development of interior-point methods, 

6: there exists a puzzling gap between the results in theory and the 

7: observations in numerical experience, i.e., algorithms with good

8: polynomial bound are not computationally efficient and algorithms

9: demonstrated efficiency in computation do not have a good or any 

10: polynomial bound. Todd raised a question in 2002: ``Can we find a 

11: theoretically and practically efficient way to reoptimize?'' This 

12: paper is an effort to close the gap. We propose two arc-search 

13: infeasible interior-point algorithms with infeasible central path

14: neighborhood wider than all existing infeasible interior-point

15: algorithms that are proved to be convergent. We show that the 

16: first algorithm is polynomial and its simplified version has a 

17: complexity bound equal to the best known

18: complexity bound for all (feasible or infeasible) interior-point

19: algorithms. We demonstrate the computational efficiency of the 

20: proposed algorithms by testing all Netlib linear programming 

21: problems in standard form and comparing the numerical results 

22: to those obtained by Mehrotra's predictor-corrector algorithm 

23: and a recently developed more efficient arc-search algorithm 

24: (the convergence of these two algorithms is unknown). We 

25: conclude that the newly proposed algorithms are not only 

26: polynomial but also computationally competitive comparing

27: to both Mehrotra's predictor-corrector algorithm and 

28: the efficient arc-search algorithm.

29:  

30: \end{abstract}