FederatedFactory: Generative One-Shot Learning for Extremely Non-IID Distributed Scenarios
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915869782704128 |
|---|---|
| author | Moleri, Andrea Internò, Christian Raza, Ali Olhofer, Markus Klindt, David Stella, Fabio Hammer, Barbara |
| author_facet | Moleri, Andrea Internò, Christian Raza, Ali Olhofer, Markus Klindt, David Stella, Fabio Hammer, Barbara |
| contents | Federated Learning (FL) enables distributed optimization without compromising data sovereignty. Yet, where local label distributions are mutually exclusive, standard weight aggregation fails due to conflicting optimization trajectories. Often, FL methods rely on pretrained foundation models, introducing unrealistic assumptions. We introduce FederatedFactory, a zero-dependency framework that inverts the unit of federation from discriminative parameters to generative priors. By exchanging generative modules in a single communication round, our architecture supports ex nihilo synthesis of universally class balanced datasets, eliminating gradient conflict and external prior bias entirely. Evaluations across diverse medical imagery benchmarks, including MedMNIST and ISIC2019, demonstrate that our approach recovers centralized upper-bound performance. Under pathological heterogeneity, it lifts baseline accuracy from a collapsed 11.36% to 90.57% on CIFAR-10 and restores ISIC2019 AUROC to 90.57%. Additionally, this framework facilitates exact modular unlearning through the deterministic deletion of specific generative modules. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_16370 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | FederatedFactory: Generative One-Shot Learning for Extremely Non-IID Distributed Scenarios Moleri, Andrea Internò, Christian Raza, Ali Olhofer, Markus Klindt, David Stella, Fabio Hammer, Barbara Machine Learning Artificial Intelligence Federated Learning (FL) enables distributed optimization without compromising data sovereignty. Yet, where local label distributions are mutually exclusive, standard weight aggregation fails due to conflicting optimization trajectories. Often, FL methods rely on pretrained foundation models, introducing unrealistic assumptions. We introduce FederatedFactory, a zero-dependency framework that inverts the unit of federation from discriminative parameters to generative priors. By exchanging generative modules in a single communication round, our architecture supports ex nihilo synthesis of universally class balanced datasets, eliminating gradient conflict and external prior bias entirely. Evaluations across diverse medical imagery benchmarks, including MedMNIST and ISIC2019, demonstrate that our approach recovers centralized upper-bound performance. Under pathological heterogeneity, it lifts baseline accuracy from a collapsed 11.36% to 90.57% on CIFAR-10 and restores ISIC2019 AUROC to 90.57%. Additionally, this framework facilitates exact modular unlearning through the deterministic deletion of specific generative modules. |
| title | FederatedFactory: Generative One-Shot Learning for Extremely Non-IID Distributed Scenarios |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2603.16370 |