Federated Coreset Subsampling for Distributed Memory Banks

Jason Witzemann; Daniel Pröll; Sebastian Hegenbart

cs.DCOct 12, 2026

Federated Coreset Subsampling for Distributed Memory Banks

Authors: Jason Witzemann, Daniel Pröll, Sebastian Hegenbart

Organizations: Digital Factory Vorarlberg GmbH, Dornbirn, Austria · University of Salzburg, Salzburg, Austria · Vorarlberg University of Applied Sciences, Dornbirn, Austria

Abstract

In real-world scenarios, industrial processes can span multiple factories and goods are often produced at multiple sites, leading to a shared yet non-identical distribution of data across locations. To address this, we propose two federated approaches for i) combining existing coresets (Double Subsampling) and ii) creating new coresets (Federated Subsampling) in a distributed anomaly detection setting. In particular, we adapt the memory bank based PatchCore algorithm to a federated setting, resulting in increased cold-start performance, resilience to domain- and sensor-shift, and resilience to unreliable actor participation, a common problem in classical, network based federated learning strategies.

Figures & tables

Method	Total	Bottle	Cable	Capsule	Carpet	Grid	Hazelnut	Leather	Metal Nut	Pill	Screw	Tile	Toothbrush	Transistor	Wood	Zipper
PatchCore	95.9	98.1	96.0	98.2	98.7	95.2	97.8	99.1	94.8	98.3	90.2	94.6	98.0	89.9	92.8	97.5
Federated Subsampling	95.9	98.1	96.0	98.2	98.7	95.2	97.8	99.1	94.8	98.3	90.2	94.6	98.0	89.9	92.8	97.5
Double Subsampling	95.8	98.0	95.8	97.9	98.6	95.3	97.7	99.1	94.7	98.0	88.1	94.8	98.0	90.6	92.7	97.5

Table 1: Comparison of Pixel-level AUROC of PatchCore, Double Subsampling, and Federated Subsampling across MVTec-AD categories

Figure 1: Federated Subsampling and per-actor pixel-level AUROC comparison with increasing training sample size

Panel of Figure 2 — Figure 2: Pixel-level AUROC comparison after using extreme and severe non-IID data distributions

Method	3 Actors	4 Actors	5 Actors
Image-Level
Baseline	$97.46\pm 4.57$	$97.46\pm 4.57$	$97.46\pm 4.57$
Random	$97.93\pm 3.93$	$97.83\pm 3.97$	$97.89\pm 3.46$
Skewed	$97.71\pm 4.13$	$97.95\pm 3.40$	$97.94\pm 3.83$
Worst	$97.41\pm 4.84$	$97.01\pm 5.79$	$96.65\pm 5.68$
Pixel-Level

Table 2: Mean AUROC

\pm

stdev across all categories of the MVTec AD dataset for different participation settings

Coreset Selection

Coreset selection chooses compact subsets that preserve a dataset’s learning, evaluation, or optimization value while reducing cost.

See all 58 papers

SCOPE: Scalable Training-Free Semantic Coreset Selection for Long-Tailed Federated Learning
Mar 13, 2026Md Anwar Hossen, Nathan R. Tallent, Luanzheng Guo +1Federated Learning Communication-Efficient Distributed Training
Iowa State University, Ames, Iowa, USA · Pacific Northwest National Laboratory, Richland, WA, USA
Memory-Bounded Continuation of Greedy Sampling for Continual Anomaly Detection
Aug 15, 2026Yoon Gyo Jung, Jaewoo Park, Kuan-Chuan Peng +2Continual Anomaly Detection Anomaly Detection
Electrical and Computer Engineering Department Northeastern University Boston, MA, USA · AIVEX Inc. Seoul, South Korea · Mitsubishi Electric Research Laboratories (MERL) Cambridge, MA, USA
Federated Low-Rank Koopman Learning for Multivariate Time-Series Anomaly Detection in IoT Systems
Jul 9, 2026Tung-Anh Nguyen, Van-Phuc Bui, Anh Tuyen Le +4Koopman Operator Learning Edge ML
Faculty of Engineering and IT, University of Technology Sydney, Ultimo, NSW, 2007, Australia · Faculty of Information Technology, FPT University, Vietnam · School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Vietnam

Mar 13, 2026J/K move · Enter open · S save

SCOPE: Scalable Training-Free Semantic Coreset Selection for Long-Tailed Federated Learning

Md Anwar Hossen, Nathan R. Tallent, Luanzheng Guo, Ali Jannesary

Iowa State University, Ames, Iowa, USA · Pacific Northwest National Laboratory, Richland, WA, USA

Federated Learning Communication-Efficient Distributed Training Long-Tail Learning

Open paper AskPDF arXiv

Scientific discovery increasingly relies on federated learning over data from high-resolution instruments, and these data are strongly non-IID and long-tailed. Coreset selection could shorten training, but existing federated coreset methods either prune with local heuristics that ignore the global class distribution, or first train a warm-up model at every client and upload class centroids of its features. This makes coreset selection a distributed bottleneck that consumes GPU time at every client, grows with the model and the number of clients, and must be repeated for every architecture. We introduce SCOPE, named for Semantic Coreset using Orthogonal Projection Embeddings, a scalable, training-free federated coreset framework for long-tailed data. Each client scores its samples once with a frozen, compact vision-language model and uploads scalar statistics per class. The server merges these into a global class profile, which lets each client remove anomalies and prune redundant samples only from classes that are locally abundant but globally common. SCOPE therefore requires no warm-up training, needs a single exchange with the server, and uploads a payload that depends only on the number of classes. On GPU nodes of an HPC cluster, with four datasets, including scientific micrographs, and six baselines, SCOPE selects its coreset up to 7.7x faster and uploads up to 293x less data than the feature-based baseline. As the number of clients grows from 10 to 500, SCOPE's time before training grows by fewer seconds than any baseline's, and at 500 clients the baselines need 1.7 to 3.6x SCOPE's time. Trained on the coreset, the global model reaches 98% of the full-data accuracy in 18 to 42% fewer global rounds than full-data training, with 8 to 21% less client computation over 200 rounds.

Abstract

Coreset Selection

Explore similar work

SCOPE: Scalable Training-Free Semantic Coreset Selection for Long-Tailed Federated Learning

Memory-Bounded Continuation of Greedy Sampling for Continual Anomaly Detection

Federated Low-Rank Koopman Learning for Multivariate Time-Series Anomaly Detection in IoT Systems