1. What are the contributions mentioned in the paper "Fast sequence segmentation using log-linear models" ?
Sequence segmentation is a well-studied problem, where given a sequence of elements, an integer K, and some measure of homogeneity, the task is to split the sequence into K contiguous segments that are maximally homogeneous.. In this paper the authors study segmentations whose measure of goodness is based on log-linear models, a rich family that contains many of the standard distributions.. The authors present a theoretical result allowing us to prune many suboptimal segmentations.. The authors demonstrate empirically, that this approach can significantly reduce the computational burden of finding the optimal segmentation.
read more
2. What have the authors stated for future works in "Fast sequence segmentation using log-linear models" ?
The authors leave these studies as future work.. The authors leave studying applying Theorem 1 more generally as future work.. The authors conjecture that using these techniques not only remove the parameter but can be also used for further speedup.. While the authors are skeptical whether it is possible verify the conditions in Theorem 1 exactly, they believe that it is possible to find more conservative conditions that can be easily checked and that will imply the conditions in Theorem 1.
read more
![Fig. 1 Toy sequence and the L2 cost of a segmentation [1, k − 1], [k, 200] as a function of k. In this paper we propose a necessary condition for a segmentation to be optimal. This condition allows us to prune suboptimal segmentations, such as [1, 100], [101, 200]](/figures/fig-1-toy-sequence-and-the-l2-cost-of-a-segmentation-1-k-1-k-2u26s422.webp)
