Maximum Subarray Problem in 1D and 2D via Weighted Paths in Directed Acyclic Graphs
Abstract
The Maximum Subarray Problem was encountered by Ulf Grenander for maximum likelihood estimation in pattern analysis. We are given a vector (or matrix) of numbers, and we have to find the contiguous sub-vector (or sub-matrix) which has the maximum sum of numbers in it. Apart from the original application, the problem also arises for example in biological sequence analysis.
We present here a linear-time algorithm in one dimension which is different from the one known due to Kadane, and present a way of extending it to two dimensions. To achieve the latter, we provide a new technique, the red-blue graphs, which encodes all the contiguous sub-matrices of an m × n matrix in size O(m × n).
Domains
Computer Science [cs]
Origin : Files produced by the author(s)
Loading...