arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.23050cs.LG
收藏

潜在因素:学习深度状态空间模型

Latent Matters: Learning Deep State-Space Models

Technical University of Munich(慕尼黑工业大学);Volkswagen Group(大众汽车集团);Eötvös Loránd University(厄特沃什·罗兰大学);AWS AI Labs(AWS人工智能实验室)

Alexej Klushyn, Richard Kurle, Maximilian Soelch, Botond Cseke, Patrick van der Smagt

AI总结:

本文提出了一种约束优化框架和扩展卡尔曼变分自编码器,用于改进深度状态空间模型的动态建模和预测准确性。

摘要:

深度状态空间模型(DSSMs)通过学习观测序列数据的底层动态来实现时间预测。它们通常通过最大化证据下界来训练。然而,正如我们所展示的,这并不能保证模型实际上学习到了底层动态。因此,我们提出了一种约束优化框架,作为训练DSSMs的一般方法。在此基础上,我们引入了扩展卡尔曼变分自编码器(EKVAE),它结合了消融变分推断与经典贝叶斯过滤/平滑技术,以更准确地建模动态。我们的结果表明,约束优化框架在已建立的最先进DSSMs示例中显著提高了系统识别和预测准确性。EKVAE在预测准确性上优于先前模型,实现了在识别动态系统方面的显著成果,并且可以进一步成功地学习到将静态和动态特征解耦的状态空间表示。

英文摘要:

Deep state-space models (DSSMs) enable temporal predictions by learning the underlying dynamics of observed sequence data. They are often trained by maximising the evidence lower bound. However, as we show, this does not ensure the model actually learns the underlying dynamics. We therefore propose a constrained optimisation framework as a general approach for training DSSMs. Building upon this, we introduce the extended Kalman VAE (EKVAE), which combines amortised variational inference with classic Bayesian filtering/smoothing to model dynamics more accurately than RNN-based DSSMs. Our results show that the constrained optimisation framework significantly improves system identification and prediction accuracy on the example of established state-of-the-art DSSMs. The EKVAE outperforms previous models w.r.t. prediction accuracy, achieves remarkable results in identifying dynamical systems, and can furthermore successfully learn state-space representations where static and dynamic features are disentangled.

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。

↑