Deep One-bit Compressive Autoencoding

TL;DR

This paper introduces a hybrid model-based and data-driven approach for designing a one-bit compressive autoencoder, improving sparse signal recovery performance over existing methods.

Contribution

It proposes a novel methodology that combines mathematical modeling and data-driven learning to optimize sensing matrices and iterative algorithms for one-bit compressed sensing.

Findings

Significant performance improvement over state-of-the-art algorithms

Effective design of sensing matrices for one-bit data acquisition

Successful learning of latent parameters for iterative recovery algorithms

Abstract

Parameterized mathematical models play a central role in understanding and design of complex information systems. However, they often cannot take into account the intricate interactions innate to such systems. On the contrary, purely data-driven approaches do not need explicit mathematical models for data generation and have a wider applicability at the cost of interpretability. In this paper, we consider the design of a one-bit compressive autoencoder, and propose a novel hybrid model-based and data-driven methodology that allows us to not only design the sensing matrix for one-bit data acquisition, but also allows for learning the latent-parameters of an iterative optimization algorithm specifically designed for the problem of one-bit sparse signal recovery. Our results demonstrate a significant improvement compared to state-of-the-art model-based algorithms.