自对偶嵌入模型解拓展熵规划

A Self-Dual Embedding Method for Extended Entropy Programming

本文把拓展熵规划转化为锥最优化问题,再对该锥最优化问题构造一个锥自对偶嵌入模型,证明了锥自对偶嵌入模型的障碍函数满足自协调性,这保证了用某些内点法求解时算法是多项式时间的.这种方法的另一个优点是不需要寻找初始可行解.

This paper transform the extended entropy programming problem into conic programming, then construct a conic self-dual embedding model for the transformed conic programming, and we prove that the barrier function is self-concordant, this guarantees the algorithm is polynomial algorithm when use some interior point method to solve this problem. And another advantage is that this method does not need to find the initial feasible solution .