Attention to Warp: Deep Metric Learning for Multivariate Time Series
Attention to Warp: Deep Metric Learning for Multivariate Time Series
复制标题
DOI：
10.1007/978-3-030-86334-0_23
复制
发表时间：
2021-03
期刊：
影响因子：
--
通讯作者：
Shinnosuke Matsuo;Xiaomeng Wu;Gantugs Atarsaikhan;Akisato Kimura;K. Kashino;Brian Kenji Iwana;S. Uchida
中科院分区：
文献类型：
--
作者：
Shinnosuke Matsuo;Xiaomeng Wu;Gantugs Atarsaikhan;Akisato Kimura;K. Kashino;Brian Kenji Iwana;S. Uchida
由于时间不变性、非线性失真和识别非匹配序列的判别力之间的困难权衡，深度时间序列度量学习具有挑战性。本文提出了一种新颖的基于神经网络的方法，用于稳健且有区别的时间序列分类和验证。这种方法使参数化注意力模型适应时间扭曲，以获得更大、更自适应的时间不变性。它不仅对局部失真而且对大的全局失真都具有鲁棒性，因此即使不满足单调性、连续性和边界条件的匹配对仍然可以被成功识别。该模型的学习进一步由动态时间扭曲引导，以施加时间约束以实现稳定的训练和更高的判别能力。它可以学习通过扭曲来增强类间变异，从而可以有效地区分相似但不同的类。在确认了该方法在 Unipen 数据集上的单字母手写分类方面的良好表现后，我们通过将其与深度在线签名验证框架相结合，通过实验证明了该方法相对于以前的非参数和深度模型的优越性。
Deep time series metric learning is challenging due to the difficult trade-off betweentemporal invarianceto nonlinear distortion anddiscriminative powerin identifying non-matching sequences. This paper proposes a novel neural network-based approach for robust yet discriminative time series classification and verification. This approach adapts a parameterized attention model to time warping for greater and more adaptivetemporal invariance. It is robust against not only local but also large global distortions, so that even matching pairs that do not satisfy the monotonicity, continuity, and boundary conditions can still be successfully identified. Learning of this model is further guided by dynamic time warping to impose temporal constraints for stabilized training and higherdiscriminative power. It can learn to augment the inter-class variation through warping, so that similar but different classes can be effectively distinguished. We experimentally demonstrate the superiority of the proposed approach over previous non-parametric and deep models by combining it with a deep online signature verification framework, after confirming its promising behavior in single-letter handwriting classification on the Unipen dataset.