Summary
Federated learning usually combines data from different places into one average model, which can ignore smaller groups with different data. The authors propose Fed-Universe, a new way to combine learning by actively negotiating between different data sources instead of just averaging. Their method looks at meaning (semantics), shape and space (geometry), relationships (topology), and human behavior to keep unique contributions from minorities. They tested this system in a simulated healthcare network and showed it helped a small minority group improve just as much as larger groups. This approach could make federated learning fairer and more personalized.
What this means in practice
- •For hospital data teams: Improve fairness in federated clinical models by preserving minority hospital data contributions during decentralized training.
- •For enterprise ai developers: Build federated AI systems that actively negotiate model updates to address data heterogeneity across multiple organizations.
Abstract
Conventional federated learners implicitly optimize for global means, creating a "one-size-fits-all" paradigm that inevitably suppresses the minority under cross-site heterogeneity. To bridge this critical gap, we present Fed-Universe, a generalizable Semantic-Geometric-Topological-Human (S-G-T-H) architecture that transforms passive averaging into an active negotiation for decentralized alignment. The S-layer deploys an edge-capable LLM as a semantic surrogate to translate heterogeneous inputs into standardized patient profile prompts. The G-layer enforces geometric quality assurance (GQA) via a mapping f: Theta -> Q from the statistical parameter space to a physically interpretable quality domain, employing a cosine similarity gate to retain minority updates based on physical geometry rather than statistical proximity. The T-layer executes topological Pareto control, identifying a sensitivity-based knee point (lambda_k) to balance multi-objective aggregation without majority degradation. Finally, the H-layer operationalizes a cognitive twin dynamic to mirror the user's real-time psychological state, dynamically adapting human-computer interaction modes via active intent alignment. To ground this theoretical architecture empirically, we validated the S-G-T core on a binational synthetic clinical network simulation. This proof-of-concept demonstrated a "zero-sum escape": a minority node (<3% data volume) achieved utility parity (loss reduction from 0.857 to 0.340) alongside dominant hubs. Our Fed-Universe framework is being expanded along two focused pillars: institutional multimodal alignment (Fed-Ultra) and individual cognitive intent alignment (Fed-Human).