Adaptive sampling using variational autoencoder and reinforcement learning

TL;DR

This paper introduces an adaptive sampling method combining variational autoencoders and reinforcement learning to improve compressed sensing by sequentially selecting measurements, leading to better reconstruction quality.

Contribution

It presents a novel adaptive sparse sensing framework that integrates deep generative priors with reinforcement learning for sequential measurement selection.

Findings

Outperforms traditional compressed sensing methods

Achieves higher reconstruction accuracy

Demonstrates effectiveness over existing generative model-based approaches

Abstract

Compressed sensing enables sparse sampling but relies on generic bases and random measurements, limiting efficiency and reconstruction quality. Optimal sensor placement uses historcal data to design tailored sampling patterns, yet its fixed, linear bases cannot adapt to nonlinear or sample-specific variations. Generative model-based compressed sensing improves reconstruction using deep generative priors but still employs suboptimal random sampling. We propose an adaptive sparse sensing framework that couples a variational autoencoder prior with reinforcement learning to select measurements sequentially. Experiments show that this approach outperforms CS, OSP, and Generative model-based reconstruction from sparse measurements.