An Inclusive and Lightweight Approach to Federated Continual Learning for Cultural Heritage

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Ioannis Theologitis, Debin Meng, Stylianos Eleftheriadis, Vasileios Lolis, Konstantinos Votis
• Venue: arXiv: cs.LG
• Published: 2026-08-20
• Collected: 2026-08-21
• URL: http://arxiv.org/abs/2608.20038v1
• PDF: https://arxiv.org/pdf/2608.20038v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

The authors propose a lightweight approach to Federated Continual Learning (FCL) called FedCurv-DR for supporting artificial intelligence in cultural heritage. This method is designed to learn from distributed and sequential data without sharing raw collections, reducing forgetting and balancing performance, fairness, and energy efficiency.

What the Document Says

• Artificial intelligence can support cultural heritage and digital humanities
• Cultural heritage data are often distributed across institutions with ownership and access restrictions
• Federated Continual Learning (FCL) is suitable for this setting as it enables models to learn from distributed and sequential data without sharing raw collections
• The authors propose FedCurv-DR, a lightweight FCL strategy that accumulates parameter-importance estimates across clients and experiences to protect learned knowledge
• Updates of these estimates are made only at fixed intervals to minimize communication and computation overhead
• Evaluation is done in a continual learning scenario using the WikiArt image dataset for genre classification with evolving styles
• Results show that FedCurv-DR reduces forgetting, balances performance, fairness, and energy efficiency

Key Concepts Mentioned

• artificial intelligence
• cultural heritage
• federated continual learning
• FedCurv-DR
• parameter-importance estimates
• distributed data
• sequential data

Claims Made by the Document

• FCL is well suited to cultural heritage data
• FedCurv-DR reduces forgetting and balances performance, fairness, and energy efficiency

Evidence Used

• The paper discusses the challenges of cultural heritage data distribution and evolving over time
• The authors describe FedCurv-DR's methodology and its ability to protect learned knowledge by updating parameter-importance estimates at fixed intervals

Methods / Approach

• Federated Continual Learning (FCL)
• proposing FedCurv-DR with accumulation of parameter-importance estimates
• evaluation in a continual learning scenario using WikiArt image dataset

Limitations Stated by the Authors

Not stated in source

Limitations Noted by the Agent

Agent interpretation — not from the source.

The evaluation is done on a specific continual learning scenario and with a specific dataset. The generalizability to other scenarios or datasets may vary.

Open Questions

• What is the impact of FedCurv-DR in other cultural heritage data types and scenarios?
• How does FedCurv-DR perform compared to other FCL strategies?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The study focuses on image genre classification with evolving styles. It may be beneficial to investigate other applications of FedCurv-DR in cultural heritage.

Raw Abstract / Excerpt

Artificial intelligence can support cultural heritage and digital humanities through large-scale retrieval and analysis of digitized collections. However, cultural heritage data are often distributed across institutions, constrained by ownership and access restrictions, and continuously evolving over time. Federated Continual Learning (FCL) is well suited to this setting, as it enables models to learn from distributed and sequential data without sharing raw collections. In this paper, we propose FedCurv-DR, a lightweight, regularisation-based FCL strategy. The method accumulates parameter-importance estimates across clients and experiences to protect learned knowledge, while updating them only at fixed intervals to minimize communication and computation overhead. We evaluate FedCurv-DR in a continual learning scenario using the WikiArt image dataset for genre classification with evolving styles, reporting performance, energy, and fairness metrics. Our results show that FedCurv- DR reduces forgetting and balances performance, fairness, and energy efficiency for sustainable AI in cultural heritage.