2 results for "federated learning"
As the Artemis and Moon-to-Mars programs advance toward a sustained lunar presence, semi-autonomous surface assets must perceive terrain onboard while sharing a scarce, thermally constrained link budget. Federated learning (FL) fits this setting: assets improve a shared perception model by exchanging updates rather than raw imagery. Existing FL studies, however, assume near-persistent connectivity and evaluate on synthetically partitioned data. This paper reports what is, to the author’s knowledge, the first evaluation of federated terrain segmentation on real planetary rover imagery under a physically motivated model of intermittent, thermally-gated connectivity. A naturally non-independent-and-identically-distributed (non-IID) three-client federation is built from the AI4Mars Curiosity dataset and the synthetic Artificial Lunar Landscape dataset; its heterogeneity is measured, not assumed (mean pairwise Jensen-Shannon divergence 0.42). A SegFormer-B0 model is trained with FedAvg while an aggregation probability pagg and a thermal duty cycle gate every round. Over 113 runs with up to five seeds per condition, accuracy declines monotonically with aggregation scarcity: federation recovers 88.8% of the gap between local-only (0.363 mIoU) and centralized (0.897) training at pagg = 1, 78.6% at 0.3, and 5.9% at 0.05. The deficit scales as paggb with b = −0.650 (95% CI [−0.755, −0.545]), steeper than the optimization-limited 1/√pagg reference and consistent with heterogeneity-driven client drift. Neither a proximal term nor server momentum recovers the scarce regime, and the temporal structure of outages, not the mean rate alone, governs predictability. Uplink volume is reported alongside accuracy at every operating point.
Federated learning (FL) lets institutions train shared models without moving raw data, attractive for clinical electroencephalography (EEG), where cohorts are small and privacy-sensitive. Two obstacles persist: clinical sites record EEG with different electrode montages, so input feature spaces differ and standard aggregation fails; and because one recording yields thousands of epochs, splitting epochs rather than subjects leaks subject identity and inflates accuracy. We present MontageFL, a montage-agnostic pipeline that maps any electrode configuration to a fixed 88-dimensional feature vector through region-based aggregation of spectral and covariance features, letting a 19-channel and an 11-channel site train together. Benchmarking centralized, local, FedAvg, and FedProx training for Alzheimer’s disease versus healthy-control classification (Miltiadous et al. dataset; 65 subjects) under strictly subject-level, leakage-free evaluation, subject-level accuracy reached approximately 74% (19-channel) and 89% (11-channel), with no detected difference between federated and local models (FedProx versus local, pp; 95% CI ), consistent with parity but unable to exclude a modest effect given the small cohort. Evaluated with epoch-level splitting, the same pipeline scored 15 to 30 points higher, illustrating leakage-driven inflation. Because montage heterogeneity was simulated by channel subsetting within one acquisition system, validation on independent recording hardware remains future work.