Ensemble Kalman Variational Objectives: Nonlinear Latent Trajectory Inference with A Hybrid of Variational Inference and Ensemble Kalman Filter

Abstract

Variational Inference (VI) combined with Bayesian nonlinear filtering produces the stateof-the-art results for latent trajectory inference. A body of recent works focused on Sequential Monte Carlo (SMC) and its expansion, e.g., Forward Filtering Backward Simulation (FFBSi). These studies achieved a great success, however, remain a serious problem for particle degeneracy. In this paper, we propose Ensemble Kalman Objectives (EnKOs), the hybrid method of VI and Ensemble Kalman Filter (EnKF), to infer the State Space Models (SSMs). Unlike the SMC based methods, the our proposed method can identify the latent dynamics given fewer particles because of its rich particle diversity. We demonstrate that EnKOs outperform the SMC based methods in terms of predictive ability for three benchmark nonlinear dynamics systems tasks.

Deep Neural Networks and End-to-End Learning for Audio Compression

“…Essentially, an end-to-end compression architecture aims to map an audio signal into a more concise portable representation. This characteristic is shared with deep generative models, targeting capturing critical aspects of data to generate new data instances (Ishizone et al [2020], Nistal et al [2021], Dhariwal et al [2020], Dieleman et al [2018]). A notable example of this approach consists of a combination of Variational Autoencoders (VAEs) and Recurrent Neural Networks (RNNs) (Hahn et al [2018]).…”