Prior-aware and Context-guided Group Sampling for Active Probabilistic Subsampling

Jul 16, 2026

Upload by Eden

Subsampling reduces measurement count, lowering data processing and transfer overhead while shortening acquisition time for many real-world applications. Active Deep Probabilistic Subsampling (A-DPS) jointly learns sampling patterns and downstream task models, but it underuses dataset priors and depends on top-1 sampling, which can limit optimization. PGA-DPS integrates deterministic prior-informed sampling from training data and group-based top-k sampling, with theoretical support for improved optimization. Experiments on classification, reconstruction, and segmentation show consistent gains over A-DPS, DPS, and other methods.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

Introduction

Related Work

arXiv :2607 .07083v1 [ ee ss . SY] 8 Jul 2026 PRIOR-AWARE AND CONTEXT-GUIDED GROUP SAMPLING FOR ACTIVE PROBABILISTIC SUBSAMPLING Beomgu Kang, Hyunseok Seo∗ Department of Artificial Intelligence, Korea University, Seoul, Republic of Korea [bgkang1@korea.ac.kr](bgkang1@korea.ac.kr) , [seoh@korea.ac.kr](seoh@korea.ac.kr) ABSTRACT Subsampling significantly reduces the number of measurements, thereby streamlining data processing and transfer overhead, and shortening acquisition time across diverse real-world applications. The recently introduced Active Deep Probabilistic Subsampling (A-DPS) approach jointly optimizes both the subsampling pattern and the downstream task model, enabling instance- and subject-specific sampling trajectories and effective adaptation to new data at inference time. However, this approach does not fully leverage valuable dataset priors and relies ontop-1 sampling, which can impede the optimization process. Herein, we enhance A-DPS by integrating a deterministic (fixed) prior-informed sampling pattern derived from the training dataset, along with group-based sampling via top-k sampling, to achieve more robust optimization—a method we call Prior-aware and context-guided Group-based Active DPS (PGA-DPS) . We also provide a theoretical analysis supporting improved optimization via group sampling, and validate this with empirical results. We evaluated PGA-DPS on three tasks: classification, image reconstruction, and segmentation, using the MNIST, CIFAR-10, fastMRI knee, and hyperspectral AeroRIT datasets, respectively. In every case, PGA-DPS outperformed A-DPS, DPS, and all other sampling methods. Our code is available at [https://github.com/B9Kang/PGADPS](https://github.com/B9Kang/PGADPS). 1 INTRODUCTION Modern technologies generate massive datasets that can require lengthy acquisition time and hinder real-time onboard processing. Many real-world applications, including Magnetic Resonance Imaging (MRI) (Ye, 2019), computed tomography (CT) imaging (Chen et al., 2008), ultrasound imaging (Huijben et al., 2020b), digital micromirror device (Baraniuk, 2007), seismic surveying (Herrmann et al., 2012), and hyperspectral imaging (Sun & Du, 2019), highlight the importance of reducing imaging data volume. Therefore, strategic sampling not only reduces data volume and preserves essential information for efficient transfer and processing but also accelerates acquisition, a critical factor in medical imaging. Compressed sensing (CS) was developed as a subsampling strategy to overcome the NyquistShannon limits on the sampling rates required for perfect signal reconstruction (Donoho, 2006 ; Eldar & Kutyniok, 2012) . CS has been widely adopted across various applications and has demonstrated significant impact (Lustig et al., 2007 ; Baraniuk & Steeghs, 2007 ; Yu & Wang, 2009 ; Mart´ınet al., 2014 ; Lorintiu et al., 2015 ; Han et al., 2016) . Although CS exploits the inherent signal structures, such as sparsity, it does not take into account the information relevant to the downstream task during the sampling process. Bridging the gap between modality-and task-specific knowledge and the sampling process has proven challenging. Recently, subsampling techniques customized for specific data distributions and downstream tasks have been proposed with advances in deep learning, offering learned yet fixed sampling patterns (Huijben et al., 2020a ; Weiss et al., 2020 ; Shen et al., 2020 ; Sherry et al., 2020 ; Zhang et al., 2020 ; Aggarwal & Jacob, 2020 ; Mou et al., 2021 ; Yang et al., 2025) . These approaches optimize a sampling pattern based on the average data distribution in the training set, which may not provide optimal results for individual instances. To address this limitation, active sampling was introduced, ∗ Corresponding author where new points are adaptively selected based on previously acquired samples, iterating until the required target number of samples is obtained (Zhang et al., 2019 ; Jin et

FAQs

On which tasks and datasets was PGA-DPS evaluated?

It was evaluated on classification, image reconstruction, and segmentation using MNIST, CIFAR-10, fastMRI knee, and hyperspectral AeroRIT datasets, outperforming A-DPS, DPS, and other sampling methods.

Jul 16, 2026

Upload by Eden