We correct a paper previously submitted to CoRR. That paper claimed that the algorithm there described was provably of linear time complexity in the average case. The alleged proof of that statement contained an error, being based on an invalid assumption, and is invalid. In this paper we present both experimental and analytical evidence that the time complexity is of order $N^2$ in the average case, where $N$ is the total length of the merged sequences.
تحميل البحث