Semi-Supervised Federated Learning
A training paradigm that combines semi-supervised learning with federated learning. It leverages a small amount of labeled data on the server side together with large volumes of unlabeled data on client devices to collaboratively train models, while preserving data privacy across participants.
Timeline (last 90 days)
Semi-supervised federated learning (SSFL) trains an initial model using a small amount of labeled seed data on the server side, then uses a teacher model to generate pseudo-labels for large volumes of unlabeled client data, thereby optimizing the model without uploading raw audio.
The combination of online pseudo-labels and server update stabilization can serve as a reference for semi-supervised federated learning in other sequence generation tasks such as machine translation and sequence labeling.
The original material this article is based on is only a solution abstract; specific experimental data, teacher model selection details, and quantified results have not yet been fully disclosed.
All Facts (3)
Semi-supervised federated learning (SSFL) trains an initial model using a small amount of labeled seed data on the server side, then uses a teacher model to generate pseudo-labels for large volumes of unlabeled client data, thereby optimizing the model without uploading raw audio.
50%UnverifiedThe combination of online pseudo-labels and server update stabilization can serve as a reference for semi-supervised federated learning in other sequence generation tasks such as machine translation and sequence labeling.
50%UnverifiedThe original material this article is based on is only a solution abstract; specific experimental data, teacher model selection details, and quantified results have not yet been fully disclosed.
50%