Differentially Private Federated Class-Incremental Learning for EEG Mental State Classification
Organizations: Institute of Industrial Automation and Software Engineering University of Stuttgart, Pfaffenwaldring 47, 70550 Stuttgart, Germany
Abstract
Federated learning (FL) under differential privacy (DP) is typically evaluated on large client cohorts, where subsampling and cross-client averaging amplify privacy and smooth injected noise. Personal biosignal applications such as wearable EEG instead involve small federations of a handful of subjects, a regime whose interaction with formal client-level DP is largely uncharacterized. We study differentially private federated class-incremental learning in this small-cohort regime, combining a public EEG dataset with a custom four-channel MUSE 2 recording set into a K=5-client federation, and give a masked per-class training protocol that avoids degenerate gradients and missing negative supervision under class-incremental streaming. We propose FedEEG, combining Synaptic-Intelligence consolidation with a lagged adaptive FedProx term, and show it reduces catastrophic forgetting relative to FedAvg, FedProx, and a federated Learning-without-Forgetting baseline; on-device balanced-batch replay reduces mean forgetting by a further 56.7%. A class-scoping diagnosis yields accuracy with 6 seeds and 2.02 times chance. We also introduce head-scoped client-level DP, which protects the classification head and alone reduces the achievable by 200 times at a matched round budget; combined with fewer communication rounds it reaches with utility not statistically distinguishable from non-private training.
Figures & tables
| Parameter | Value |
|---|---|
| Window / stride | 256 / 128 samples (50% overlap) |
| Rounds/increment, local epochs (main) | 3, 2 |
| Batch size, optimizer | 64, Adam ( , reset/round) |
| Aggregation weighting | -weighted (Sec. V-A ) |
| SI: | 25.0, 1.0, |
| Prox: | 5.0, 1.0 |
| Configuration | Classes | Seeds | ACC (chance ) |
|---|---|---|---|
| 5-class (full) | all 5 | 4 | (2.33 ) |
| 3-class (shared) | neu., rel., conc. | 4 | (1.40 ) |
| 4-class (drop alerted) | neu., rel., conc., drow. | 6 | ( 2.02 ) |
| Method | ACC | BWT | FM |
|---|---|---|---|
| FedAvg (CE) | |||
| CE + Prox | |||
| Federated LwF [ 16 ] | |||
| CE + Prox + SI | |||
| + Balanced Replay ( V-B1 ) |
| Configuration | ACC | BWT | FM |
|---|---|---|---|
| Non-DP, reduced-round (3 rd.) | |||
| DP, head-scoped, | |||
| Non-DP, original (9 rd.) |
Federated Learning
Federated learning trains shared models across data holders without requiring them to pool their raw data.
See all 1,027 papers