论文标题
马尔可夫高斯流程变异自动编码器
Markovian Gaussian Process Variational Autoencoders
论文作者
论文摘要
对于许多高维时间序列建模问题，已成功考虑了顺序VAE，许多变体模型依赖于离散的时间机制，例如复发性神经网络（RNN）。另一方面，连续时间方法最近引起了吸引力，尤其是在不规则采样的时间序列的背景下，它们可以比离散时间方法更好地处理数据。这样的类是高斯工艺变异自动编码器（GPVAE），其中VAE先验设置为高斯过程（GP）。但是，GPVAE的主要局限性是它将立方计算成本作为GPS继承，使其对从事的人没有吸引力。在这项工作中，我们利用马尔可夫GP的等效离散状态空间表示，通过卡尔曼过滤和平滑来实现线性时间GPVAE训练。对于我们的模型Markovian GPVAE（MGPVAE），我们在各种高维时和时空任务上展示了我们的方法与现有方法相比的性能优惠，同时计算高度可扩展性。
Sequential VAEs have been successfully considered for many high-dimensional time series modelling problems, with many variant models relying on discrete-time mechanisms such as recurrent neural networks (RNNs). On the other hand, continuous-time methods have recently gained attraction, especially in the context of irregularly-sampled time series, where they can better handle the data than discrete-time methods. One such class are Gaussian process variational autoencoders (GPVAEs), where the VAE prior is set as a Gaussian process (GP). However, a major limitation of GPVAEs is that it inherits the cubic computational cost as GPs, making it unattractive to practioners. In this work, we leverage the equivalent discrete state space representation of Markovian GPs to enable linear time GPVAE training via Kalman filtering and smoothing. For our model, Markovian GPVAE (MGPVAE), we show on a variety of high-dimensional temporal and spatiotemporal tasks that our method performs favourably compared to existing approaches whilst being computationally highly scalable.
