Masked Autoencoders are Efficient Continual Federated Learners

Subarnaduti Paul, Lars-Joel Frey, Roshni Ramanna Kamath, Kristian Kersting, Martin Mundt
Proceedings of The 3rd Conference on Lifelong Learning Agents, PMLR 274:70-85, 2025.

Abstract

Machine learning is typically framed from a perspective of i.i.d., and more importantly, isolated data. In parts, federated learning lifts this assumption, as it sets out to solve the real-world challenge of collaboratively learning a shared model from data distributed across clients. However, motivated primarily by privacy and computational constraints, the fact that data may change, distributions drift, or even tasks advance individually on clients, is seldom taken into account. The field of continual learning addresses this separate challenge and first steps have recently been taken to leverage synergies in distributed settings of a purely supervised nature. Motivated by these prior works, we posit that such federated continual learning should be grounded in unsupervised learning of representations that are shared across clients; in the loose spirit of how humans can indirectly leverage others’ experience without exposure to a specific task. For this purpose, we demonstrate that masked autoencoders for distribution estimation are particularly amenable to this setup. Specifically, their masking strategy can be seamlessly integrated with task attention mechanisms to enable selective knowledge transfer between clients. We empirically corroborate the latter statement through several continual federated scenarios on both image and binary datasets.

Cite this Paper


BibTeX
@InProceedings{pmlr-v274-paul25a, title = {Masked Autoencoders are Efficient Continual Federated Learners}, author = {Paul, Subarnaduti and Frey, Lars-Joel and Kamath, Roshni Ramanna and Kersting, Kristian and Mundt, Martin}, booktitle = {Proceedings of The 3rd Conference on Lifelong Learning Agents}, pages = {70--85}, year = {2025}, editor = {Lomonaco, Vincenzo and Melacci, Stefano and Tuytelaars, Tinne and Chandar, Sarath and Pascanu, Razvan}, volume = {274}, series = {Proceedings of Machine Learning Research}, month = {29 Jul--01 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v274/main/assets/paul25a/paul25a.pdf}, url = {https://proceedings.mlr.press/v274/paul25a.html}, abstract = {Machine learning is typically framed from a perspective of i.i.d., and more importantly, isolated data. In parts, federated learning lifts this assumption, as it sets out to solve the real-world challenge of collaboratively learning a shared model from data distributed across clients. However, motivated primarily by privacy and computational constraints, the fact that data may change, distributions drift, or even tasks advance individually on clients, is seldom taken into account. The field of continual learning addresses this separate challenge and first steps have recently been taken to leverage synergies in distributed settings of a purely supervised nature. Motivated by these prior works, we posit that such federated continual learning should be grounded in unsupervised learning of representations that are shared across clients; in the loose spirit of how humans can indirectly leverage others’ experience without exposure to a specific task. For this purpose, we demonstrate that masked autoencoders for distribution estimation are particularly amenable to this setup. Specifically, their masking strategy can be seamlessly integrated with task attention mechanisms to enable selective knowledge transfer between clients. We empirically corroborate the latter statement through several continual federated scenarios on both image and binary datasets.} }
Endnote
%0 Conference Paper %T Masked Autoencoders are Efficient Continual Federated Learners %A Subarnaduti Paul %A Lars-Joel Frey %A Roshni Ramanna Kamath %A Kristian Kersting %A Martin Mundt %B Proceedings of The 3rd Conference on Lifelong Learning Agents %C Proceedings of Machine Learning Research %D 2025 %E Vincenzo Lomonaco %E Stefano Melacci %E Tinne Tuytelaars %E Sarath Chandar %E Razvan Pascanu %F pmlr-v274-paul25a %I PMLR %P 70--85 %U https://proceedings.mlr.press/v274/paul25a.html %V 274 %X Machine learning is typically framed from a perspective of i.i.d., and more importantly, isolated data. In parts, federated learning lifts this assumption, as it sets out to solve the real-world challenge of collaboratively learning a shared model from data distributed across clients. However, motivated primarily by privacy and computational constraints, the fact that data may change, distributions drift, or even tasks advance individually on clients, is seldom taken into account. The field of continual learning addresses this separate challenge and first steps have recently been taken to leverage synergies in distributed settings of a purely supervised nature. Motivated by these prior works, we posit that such federated continual learning should be grounded in unsupervised learning of representations that are shared across clients; in the loose spirit of how humans can indirectly leverage others’ experience without exposure to a specific task. For this purpose, we demonstrate that masked autoencoders for distribution estimation are particularly amenable to this setup. Specifically, their masking strategy can be seamlessly integrated with task attention mechanisms to enable selective knowledge transfer between clients. We empirically corroborate the latter statement through several continual federated scenarios on both image and binary datasets.
APA
Paul, S., Frey, L., Kamath, R.R., Kersting, K. & Mundt, M.. (2025). Masked Autoencoders are Efficient Continual Federated Learners. Proceedings of The 3rd Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 274:70-85 Available from https://proceedings.mlr.press/v274/paul25a.html.

Related Material