Computing Block-Angular Karmarkar Projections with Applications to Stochastic Programming
提出一种针对块角结构线性规划（如随机线性规划）的Karmarkar算法变体，通过高效计算投影，在最坏情况下运行时间比标准算法快一个数量级，并讨论了近似和大规模问题的应用。
We present a variant of Karmarkar's algorithm for block-angular structured linear programs, such as stochastic linear programs. By computing the projection efficiently, we give a worst-case bound on the order of the running time that can be an order of magnitude better than that of Karmarkar's standard algorithm. Further implications for approximations and very large-scale problems are given.