Deep Poisson gamma dynamical systems
arXiv:1810.11209
Abstract
We develop deep Poisson-gamma dynamical systems (DPGDS) to model sequentially observed multivariate count data, improving previously proposed models by not only mining deep hierarchical latent structure from the data, but also capturing both first-order and long-range temporal dependencies. Using sophisticated but simple-to-implement data augmentation techniques, we derived closed-form Gibbs sampling update equations by first backward and upward propagating auxiliary latent counts, and then forward and downward sampling latent variables. Moreover, we develop stochastic gradient MCMC inference that is scalable to very long multivariate count time series. Experiments on both synthetic and a variety of real-world data demonstrate that the proposed model not only has excellent predictive performance, but also provides highly interpretable multilayer latent structure to represent hierarchical and temporal information propagation.
arXiv commentNeurIPS 2018
References in corpus (10)
- Continuous Time Dynamic Topic Models
- Beta-Negative Binomial Process and Poisson Factor Analysis
- Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks
- Infinite Edge Partition Models for Overlapping Community Detection and Link Prediction
- Dynamic Poisson Factorization
- WHAI: Weibull Hybrid Autoencoding Inference for Deep Topic Modeling
- Deep Temporal Sigmoid Belief Networks for Sequence Modeling
- Nonparametric Bayesian Factor Analysis for Dynamic Count Matrices
- Poisson--Gamma Dynamical Systems
- The Poisson Gamma Belief Network