Research map / Federated learning and privacy / Privacy attacks and secure learning

Scalable and Trustworthy Learning in Heterogeneous Networks

AAAI 2025 · Federated learning and privacy › Privacy attacks and secure learning · approach: Private federated aggregation · 2 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

An AAAI 2025 invited-talk abstract summarizing the author's work on scalable and trustworthy federated learning under data heterogeneity, including FedProx, q-FFL, Ditto, and TERM.

Problem

Federated learning across heterogeneous, massive, and unreliable networks faces challenges in scalability, convergence, fairness, robustness to noisy or poisoned data, and privacy, all of which are worsened by data heterogeneity.

Approach

Summarizes a line of work: a proximal term for stable federated optimization (FedProx), a tunable q-parameterized reweighted loss for fairness (q-FFL), a multi-task personalization method (Ditto) for jointly meeting accuracy, fairness and robustness, a tilted ERM objective (TERM) for outliers and distribution shift, and gradient-structure-based DP adaptive optimizers.

Key results

• FedProx is presented as the first optimization framework with theoretical guarantees under realistic heterogeneity assumptions.
• q-FFL minimizes an aggregate reweighted loss parameterized by q to upweight high-loss clients, giving a tunable fairness/utility tradeoff.
• TERM introduces a tilting hyperparameter that can provably decrease the influence of outliers, enable fairness or robustness, and improve generalization.

Limitations

• Abstract-level summary with no experimental details or numbers reported in the text
• Future research plans are only briefly described

Experimental setup

Builds on: FedProx FedDANE q-FFL Ditto TERM

Closest papers by meaning

• An Aggregation-Free Federated Learning for Tackling Data Heterogeneity (CVPR 2024 · similarity 0.88)
• Fake It Till Make It: Federated Learning with Consensus-Oriented Generation (ICLR 2024 · similarity 0.88)
• Overcoming Data and Model heterogeneities in Decentralized Federated Learning via Synthetic Anchors (ICML 2024 · similarity 0.87)
• A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.87)
• Federated Learning with Heterogeneous LLMs: Integrating Small Student Client Models with a Large Hungry Model (AAAI 2025 · similarity 0.87)
• Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning (ICML 2024 · similarity 0.87)