EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE

Abstract

Many real-life decision-making situations allow further relevant information to be acquired at a specific cost, for example, in assessing the health status of a patient we may decide to take additional measurements such as diagnostic tests or imaging scans before making a final assessment. Acquiring more relevant information enables better decision making, but may be costly. How can we trade off the desire to make good decisions by acquiring further information with the cost of performing that acquisition? To this end, we propose a principled framework, named EDDI (Efficient Dynamic Discovery of high-value Information), based on the theory of Bayesian experimental design. In EDDI, we propose a novel partial variational autoencoder (Partial VAE) to predict missing data entries problematically given any subset of the observed ones, and combine it with an acquisition function that maximizes expected information gain on a set of target variables. We show cost reduction at the same decision quality and improved decision quality at the same cost in multiple machine learning benchmarks and two real-world health-care applications.

arXiv commenticml 2019 camera-ready version

Cited by in corpus (28)

• GP-VAE: Deep Probabilistic Time Series Imputation
• not-MIWAE: Deep Generative Modelling with Missing not at Random Data
• Scalable Gaussian Process Variational Autoencoders
• Instructions and Guide for Diagnostic Questions: The NeurIPS 2020 Education Challenge
• VAEM: a Deep Generative Model for Heterogeneous Mixed Type Data
• Identifiable Generative Models for Missing Not at Random Data Imputation
• VAEs in the Presence of Missing Data
• Conservative Policy Construction Using Variational Autoencoders for Logged Data with Missing Values
• Icebreaker: Element-wise Active Information Acquisition with Bayesian Deep Latent Gaussian Model
• Active Feature Acquisition with Generative Surrogate Models
• VarFA: A Variational Factor Analysis Framework For Efficient Bayesian Learning Analytics
• A Bayesian Approach for Medical Inquiry and Disease Inference in Automated Differential Diagnosis
• Variational Selective Autoencoder: Learning from Partially-Observed Heterogeneous Data
• Learning Multimodal VAEs through Mutual Supervision
• Reinforcement Learning with Efficient Active Feature Acquisition
• Contextual HyperNetworks for Novel Feature Adaptation
• Gradient Importance Learning for Incomplete Observations
• Towards Robust Active Feature Acquisition
• DADI: Dynamic Discovery of Fair Information with Adversarial Reinforcement Learning
• On the Consistency of a Random Forest Algorithm in the Presence of Missing Entries
• Networked Time Series Prediction with Incomplete Data via Generative Adversarial Network
• Simultaneous Measurement Imputation and Outcome Prediction for Achilles Tendon Rupture Rehabilitation
• CoRGi: Content-Rich Graph Neural Networks with Attention
• Causally Constrained Data Synthesis for Private Data Release
• EMFlow: Data Imputation in Latent Space via EM and Deep Flow Models
• Medical data wrangling with sequential variational autoencoders
• Dynamic Feature Acquisition with Arbitrary Conditional Flows
• Adaptive Covariate Acquisition for Minimizing Total Cost of Classification