ICVGIP 2025 Paper ID 258 - Pooling Diverse Voices: Logarithmic Binary Fusion for Server-Side
Автор: ICVGIP2025
Загружено: 2025-12-12
Просмотров: 11
Описание:
Pooling Diverse Voices: Logarithmic Binary Fusion for Server-Side Pseudo-Labeling in Federated Learning (LFBC)
"This video presents LFBC — a federated learning framework that replaces fragile multiclass models with independent binary classifiers and fuses client predictions in log-space. By using LGM and MLO fusion, the server generates stable pseudo-labels for its unlabeled data, even when clients have heterogeneous models and non-IID distributions. LFBC improves accuracy, reduces communication, and supports incremental expansion of classes, offering a more scalable and flexible FL solution.
"
Authors:
Aditi Palit, Kalidas Yeturu
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: