Deep Kalman Filters
arXiv:1511.05121
Abstract
Kalman Filters are one of the most influential models of time-varying phenomena. They admit an intuitive probabilistic interpretation, have a simple functional form, and enjoy widespread adoption in a variety of disciplines. Motivated by recent variational methods for learning deep generative models, we introduce a unified algorithm to efficiently learn a broad spectrum of Kalman filters. Of particular interest is the use of temporal generative models for counterfactual inference. We investigate the efficacy of such models for counterfactual inference, and to that end we introduce the "Healing MNIST" dataset where long-term structure, noise and actions are applied to sequences of digits. We show the efficacy of our method for modeling this dataset. We further show how our model can be used for counterfactual inference for patients, based on electronic health record data of 8,000 patients over 4.5 years.
17 pages, 14 figures: Fixed typo in Fig. 1(b) and added reference
References in corpus (3)
Cited by in corpus (75)
- KalmanNet: Neural Network Aided Kalman Filtering for Partially Known Dynamics
- Deep Learning and its Application to LHC Physics
- Deep Learning for Physical Processes: Incorporating Prior Scientific Knowledge
- Composing graphical models with neural networks for structured representations and fast inference
- GRU-ODE-Bayes: Continuous modeling of sporadically-observed time series
- Dream to Control: Learning Behaviors by Latent Imagination
- Black box variational inference for state space models
- Backprop KF: Learning Discriminative Deterministic State Estimators
- Clinical Intervention Prediction and Understanding using Deep Networks
- Deep Variational Bayes Filters: Unsupervised Learning of State Space Models from Raw Data
- LFADS - Latent Factor Analysis via Dynamical Systems
- Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Model
- Spatio-Temporal Neural Networks for Space-Time Series Forecasting and Relations Discovery
- Physics-guided Deep Markov Models for Learning Nonlinear Dynamical Systems with Uncertainty
- Variational Temporal Deep Generative Model for Radar HRRP Target Recognition
- Learning and Querying Fast Generative Models for Reinforcement Learning
- Conjugate-Computation Variational Inference : Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models
- Linear dynamical neural population models through nonlinear embeddings
- Generative Temporal Models with Memory
- Deep Amortized Inference for Probabilistic Programs
- Deconfounding Reinforcement Learning in Observational Settings
- Creativity: Generating Diverse Questions using Variational Autoencoders
- State Space LSTM Models with Particle MCMC Inference
- GP-VAE: Deep Probabilistic Time Series Imputation
- Advances in Variational Inference
- Interpretable End-to-end Urban Autonomous Driving with Latent Deep Reinforcement Learning
- Mastering Atari with Discrete World Models
- Recurrent Attentive Neural Process for Sequential Data
- Set Functions for Time Series
- Action and Perception as Divergence Minimization
- Deep Generative Video Compression
- Deep Factors with Gaussian Processes for Forecasting
- Structured Object-Aware Physics Prediction for Video Modeling and Planning
- Learning Dynamics Model in Reinforcement Learning by Incorporating the Long Term Future
- Disentangled State Space Representations
- Contrastively Disentangled Sequential Variational Autoencoder
- Temporal Difference Variational Auto-Encoder
- Semi-Supervised Generation with Cluster-aware Generative Models
- Lyapunov-Based Reinforcement Learning State Estimator
- Learning Awareness Models
- Information Maximizing Visual Question Generation
- Recurrent Neural Networks with Stochastic Layers for Acoustic Novelty Detection
- Learning Interpretable Deep State Space Model for Probabilistic Time Series Forecasting
- Particle Smoothing Variational Objectives
- Variational Deep Learning for the Identification and Reconstruction of Chaotic and Stochastic Dynamical Systems from Noisy and Partial Observations
- ACE-NODE: Attentive Co-Evolving Neural Ordinary Differential Equations
- Action-Sufficient State Representation Learning for Control with Structural Constraints
- Variational Inference for Data-Efficient Model Learning in POMDPs
- Probabilistic Video Generation using Holistic Attribute Control
- Socially Aware Kalman Neural Networks for Trajectory Prediction
- Variational online learning of neural dynamics
- A Survey of Techniques All Classifiers Can Learn from Deep Networks: Models, Optimizations, and Regularization
- A Worrying Analysis of Probabilistic Time-series Models for Sales Forecasting
- Stochastic Sequential Neural Networks with Structured Inference
- Estimating Nonlinear Dynamics with the ConvNet Smoother
- Deep Markov Spatio-Temporal Factorization
- End-to-end Autonomous Driving Perception with Sequential Latent Representation Learning
- Estimating Granger Causality with Unobserved Confounders via Deep Latent-Variable Recurrent Neural Network
- End-To-End Semi-supervised Learning for Differentiable Particle Filters
- Energy-Inspired Models: Learning with Sampler-Induced Distributions
- Bidirectional Inference Networks: A Class of Deep Bayesian Networks for Health Profiling
- Harnessing value from data science in business: ensuring explainability and fairness of solutions
- Ensemble Kalman Variational Objectives: Nonlinear Latent Trajectory Inference with A Hybrid of Variational Inference and Ensemble Kalman Filter
- Beta DVBF: Learning State-Space Models for Control from High Dimensional Observations
- A Generative Map for Image-based Camera Localization
- Improve Agents without Retraining: Parallel Tree Search with Off-Policy Correction
- Mind the Gap when Conditioning Amortised Inference in Sequential Latent-Variable Models
- Re-examination of the Role of Latent Variables in Sequence Modeling
- Learning Dynamical Systems from Noisy Sensor Measurements using Multiple Shooting
- Dynamic Variational Autoencoders for Visual Process Modeling
- A Benchmark of Dynamical Variational Autoencoders applied to Speech Spectrogram Modeling
- Self-Supervised Inference in State-Space Models
- Variational Autoencoded Regression: High Dimensional Regression of Visual Data on Complex Manifold
- Switching Recurrent Kalman Networks
- Deep Measurement Updates for Bayes Filters