# Task-Based Analog-to-Digital Converters

arXiv: https://arxiv.org/abs/2009.14088 · HTML: https://mlsanity.com/paper/2009.14088/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 29 Sep 2020.

## Built on by (most cited first, 7 of 7)

- [Task-Oriented Sensing, Computation, and Communication Integration for Multi-Device Edge AI](https://mlsanity.com/paper/2207.00969/) (2022)
- [LoRD-Net: Unfolded Deep Detection Network with Low-Resolution Receivers](https://mlsanity.com/paper/2102.02993/) (2021)
- [Deep Task-Based Analog-to-Digital Conversion](https://mlsanity.com/paper/2201.12634/) (2022)
- [Revisiting Outage for Edge Inference Systems](https://mlsanity.com/paper/2504.03686/) (2025)
- [Design and Analysis of Hardware-limited Non-uniform Task-based Quantizers](https://mlsanity.com/paper/2208.07525/) (2022)
- [Remote Inference over Dynamic Links via Adaptive Rate Deep Task-Oriented Vector Quantization](https://mlsanity.com/paper/2501.02521/) (2025)
- [Task-Based Graph Signal Compression](https://mlsanity.com/paper/2110.12387/) (2021)

## Abstract (arXiv)

Obtaining digital representations of multivariate continuous-time (CT) signals is a challenge encountered in many signal processing systems. In practice, these signals are often acquired to extract some underlying information, i.e., for a specific task. Employing conventional task-agnostic analog-to-digital converters (ADCs), typically designed to minimize the mean squared error (MSE) in reconstructing the CT input signal, can be costly and energy-inefficient in such cases. In this work, we study task-based ADCs, which are designed to obtain a digital representation of a multivariate CT input process with the goal of recovering an underlying statistically related parameter vector, referred to as the task. The proposed system employs analog filtering, uniform sampling, and scalar uniform quantization of the input process before recovering the task vector using a linear digital recovery filter. We optimize the analog and digital filters and derive closed-form expressions for the achievable MSE in recovering the task vector from a set of analog signals when utilizing ADCs with a fixed sampling rate and amplitude resolution. Based on our derivation, we provide guidelines for designing practical acquisition systems subject to a constraint on the bit rate. Our analysis proves that the intuitive approaches of either recovering the task vector solely in digital or designing the analog filter to estimate the task vector are inferior to the proposed joint design. We then consider the recovery of a set of matched filter outputs under a rate budget. We numerically verify our theoretical observations and demonstrate that task-based ADCs substantially outperform analog matched filtering as well as applying the matched filter solely in the digital domain. [...]