arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2010.12972eess.SPcs.LG
收藏

基于自注意力的软最小代价流学习实现时间序列信号盲去交错

Blind Deinterleaving of Signals in Time Series with Self-attention Based Soft Min-cost Flow Learning

ASELSAN A. Ş.(ASELSAN公司)

Oğul Can, Yeti Z. Gürbüz, Berkin Yıldırım, A. Aydın Alatan

AI总结:

针对雷达等时间序列信号的盲去交错问题,提出将聚类等价为最小代价流,并用自注意力网络近似双层优化的下层问题,实现端到端聚类。

摘要:

我们提出一种端到端学习方法,以解决时间序列中模式的去交错问题,尤其是雷达信号。一旦存在适当代价,我们将信号聚类问题与最小代价流联系起来,视为等价问题。我们构建了一个双层优化问题,其中将最小代价流作为子问题,以便从有监督训练数据中学习此类代价。随后,我们利用基于自注意力的神经网络近似下层优化问题,并提供了一个可训练框架,将输入中的模式聚类为不同的流。我们在包含若干具有挑战性场景的大型数据集上通过大量实验评估了该方法,以证明其效率。

英文摘要:

We propose an end-to-end learning approach to address deinterleaving of patterns in time series, in particular, radar signals. We link signal clustering problem to min-cost flow as an equivalent problem once the proper costs exist. We formulate a bi-level optimization problem involving min-cost flow as a sub-problem to learn such costs from the supervised training data. We then approximate the lower level optimization problem by self-attention based neural networks and provide a trainable framework that clusters the patterns in the input as the distinct flows. We evaluate our method with extensive experiments on a large dataset with several challenging scenarios to show the efficiency.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑