---
title: Adaptive Fair Federated Learning (AFFL)
url: https://www.emergentmind.com/topics/adaptive-fair-federated-learning-affl
type: topic
---

# Adaptive Fair Federated Learning (AFFL)

Searching arXiv for recent papers on Adaptive Fair Federated Learning and closely related methods.
Adaptive Fair Federated Learning (AFFL) denotes a family of federated learning methods in which fairness is treated as an online control signal for training, rather than as a purely post hoc diagnostic. In this literature, “fairness” is not a single notion. It can mean more uniform performance across clients, group fairness with respect to sensitive attributes, fair client selection, equitable participation frequency, or preference-conditioned trade-offs between fairness and utility. Correspondingly, AFFL methods adapt different components of the federated pipeline: server-side aggregation weights, client-selection probabilities, local learning rates, Momentum terms, clipping rules, preference-conditioned personalized models, or synthetic-data augmentation mechanisms [2110.00857][2209.13678][2401.16350].

## 1. Emergence and scope of AFFL

The earliest methods in this line of work concentrated on **server-side fairness-aware aggregation**. FairFed proposed a server-side method that remains agnostic to the applied local debiasing and changes aggregation weights round by round based on a fairness gap between local and global fairness behavior [2110.00857]. FAIR-FATE extended this direction by introducing a fairness-aware aggregation update with Momentum, using a server validation set to decide which client updates are “more fair” than the current global model and a time-varying fairness coefficient $\lambda_t$ that increases over rounds [2209.13678].

Subsequent work expanded the adaptive component beyond aggregation alone. AdaFedAdam reformulated fair federated learning as a **dynamic multi-objective optimization problem** based on inverse training rates and coupled it with certainty-adaptive Adam at the server [2301.09357]. FedEBA+ combined entropy-based aggregation with model and gradient alignments in a bi-level optimization framework [2301.12407]. AdaFed constructed a **common descent direction** that decreases all clients’ losses while making higher-loss clients decrease faster [2401.04993]. FedFair$^3$ moved the adaptive intervention to **utility-aware client selection**, explicitly targeting fair client selection, equitable participation frequency, and a fair accuracy distribution across clients [2401.16350].

By 2024–2025, the AFFL umbrella had broadened further. FedAA treated aggregation as a reinforcement-learning control problem under non-IID data and Byzantine attacks [2402.05541]. FedFDP coupled fairness with differential privacy through fairness-aware gradient clipping and adaptive loss clipping [2402.16028]. PraFFL replaced a single fairness–accuracy operating point with preference-conditioned personalized models generated by client-side hypernetworks [2404.08973]. AAggFF recast fairness-aware aggregation as an online convex optimization problem with sublinear regret guarantees [2405.20821]. OTA-FFL incorporated adaptive fairness-aware aggregation into over-the-air computation [2501.03392]. AFed used conditional generation and conditional GANs to approximate a global sensitive-attribute-conditioned distribution for fair data augmentation [2501.02732]. A later medical-AI framework uses “Adaptive Fair Federated Learning (AFFL)” explicitly as a systems-level research program built around adaptive knowledge messengers, fairness-aware distillation, and curriculum-guided acceleration, though its reported gains are largely projected rather than fully established experimental outcomes [2510.06259].

This suggests that AFFL is best understood not as a single algorithm, but as a research area centered on **adaptive fairness control under heterogeneity**.

## 2. Fairness objectives and formal problem settings

Most AFFL papers begin from the standard federated objective
\[
\min_{\theta} f(\theta)=\sum_{k \in K} \frac{n_k}{n} G_k(\theta)
\]
or closely related weighted formulations such as
\[
\mathbf{w}^* \triangleq \arg \min F(\mathbf{w}), \qquad F(\mathbf{w}) \triangleq \sum_{i=1}^N p_i F_i(\mathbf{w}).
\]
The departure from ordinary FL is that fairness enters either the objective itself, the aggregation rule, or the update geometry [2209.13678][2402.16028].

A major branch of AFFL defines fairness as **uniformity of client performance**. FedAA states that a model is fairer if
\[
\mathrm{std}\{F_k(w_1)\}_{k\in[N]} < \mathrm{std}\{F_k(w_2)\}_{k\in[N]},
\]
and OTA-FFL adopts the same standard-deviation proxy over client test performance [2402.05541][2501.03392]. FedFDP uses a weighted variance of client losses,
\[
\Psi(\mathbf{w}) \triangleq  \sum_{i=1}^N p_i \left(F_i(\mathbf{w})-F(\mathbf{w})\right)^2,
\]
with smaller $\Psi$ meaning better fairness [2402.16028]. FeDABoost measures fairness by the variance of client performance $\mathrm{Var}(\varphi(\mathcal{M}))$, while AdaFed emphasizes a common descent direction that is descent for every client and decreases higher-loss clients faster [2510.02914][2401.04993].

A second branch targets **group fairness**. FAIR-FATE uses Statistical Parity, Equality of Opportunity, and Equalized Odds, all with ideal value $1$:
\[
SP = \frac{P[\hat{Y}=1 \mid S=0]}{P[\hat{Y}=1 \mid S=1]}, \qquad
EO = \frac{P[\hat{Y}=1 \mid S=0, Y=1]}{P[\hat{Y}=1 \mid S=1, Y=1]}.
\]
FairFed works with Equal Opportunity Difference and Statistical Parity Difference,
\[
EOD = \Pr(\hat{Y}=1 \mid A=0, Y=1) - \Pr(\hat{Y}=1 \mid A=1, Y=1),
\]
\[
SPD = \Pr(\hat{Y}=1 \mid A=0) - \Pr(\hat{Y}=1 \mid A=1),
\]
and distinguishes sharply between **global fairness** over the federated population and **local fairness** on each client’s own distribution [2209.13678][2110.00857]. PraFFL uses demographic parity disparity,
\[
\Delta {DP}(\mathcal{D}, \bm{\theta})= \left| \mathbb{E}_{(\bm{x}, y)\in \mathcal{D}\mid a=0}\mathbb{I}_{\hat{y}=1} - \mathbb{E}_{(\bm{x}, y)\in \mathcal{D}\mid a=1}\mathbb{I}_{\hat{y}=1} \right|,
\]
and AFed measures discrimination by $\Delta \textrm{DP}$ as well [2404.08973][2501.02732].

FedFair$^3$ adds a third perspective by making fairness explicitly **multi-dimensional**: client selection fairness, participation-frequency fairness, and fairness in the distribution of client accuracies. Its formulation starts from
\[
\min_w F(w) = \sum \limits_{i=1}^N  {p_i}\alpha_i F_i(w),
\]
with the fairness argument tied not only to model outcomes but also to who participates and how often [2401.16350].

A common misconception is that AFFL always refers to demographic fairness. The literature does not support that restriction. Some methods are explicitly demographic, some are explicitly client-performance-oriented, and some combine participation or contribution fairness with outcome fairness.

## 3. Core adaptive mechanisms

The central technical idea in AFFL is that fairness is enforced by **changing the training dynamics during federation**. The mechanisms differ substantially across methods.

FairFed performs what it describes as **closed-loop fairness control**. Its per-client gap is
\[
\Delta_k^t = \begin{cases} \left|Acc_k^t - \overline{Acc^t}\right| & \text{if } F_k^t \text{ is undefined}, \\
\left|F_{global}^t - F_k^t\right| & \text{otherwise},
\end{cases}
\]
and it updates an unnormalized weight accumulator via
\[
\bar{\omega}_k^t = \bar{\omega}_k^{t-1} - \beta\left(\Delta_k - \frac{1}{K}\sum_{i=1}^K \Delta_i\right),
\qquad
\omega_k^t = \frac{\bar{\omega}_k^t}{\sum_{i=1}^K \bar{\omega}_i^t}.
\]
Larger $\beta$ strengthens fairness influence and can reduce accuracy [2110.00857].

FAIR-FATE separates normal and fairness-oriented updates. It computes the standard data-size-weighted update $\alpha_N$, a fairness-aware aggregation update $\alpha_F$ over clients whose local models are more fair than the current global model, smooths $\alpha_F$ with a decaying Momentum term
\[
v_{t+1} = \beta_t v_t + (1 - \beta_t)\alpha_F,
\]
and combines it with the ordinary update using
\[
\theta_{t+1} = \theta_t + \lambda_t v_{t+1} + (1 - \lambda_t)\alpha_N.
\]
Here $\lambda_t$ increases over rounds and $\beta_t$ decays [2209.13678].

AdaFed uses a different route. It orthogonalizes client pseudo-gradients, scales them by loss, and computes a minimum-norm vector in the convex hull of the orthogonalized gradients. The resulting server direction satisfies the property
\[
\nabla f_k(\theta_t)\cdot \mathfrak d_t \propto |f_k(\theta_t)|^\gamma,
\]
so all clients descend, but higher-loss clients descend faster [2401.04993].

FedEBA+ derives adaptive aggregation weights from a maximum-entropy principle:
\[
p_i=\frac{\exp [F_i(x) / \tau)]}{\sum_{i=1}^N \exp [F_i(x) / \tau]},
\]
then blends aggregated gradients with an “ideal” aligned direction to balance fairness and global performance [2301.12407]. AAggFF instead treats aggregation as a sequential decision process and updates the mixing vector online using an exponentiated-gradient-style rule
\[
p_i^{(t+1)} \propto p_i^{(t)}\exp\left(\frac{r_i^{(t)}}{\eta}\right),
\]
or an ONS-style update in the cross-silo regime, depending on the setting [2405.20821]. OTA-FFL computes round-wise adaptive weights through a modified Chebyshev scalarization constrained around the FedAvg weights, with $\epsilon$ controlling the fairness–accuracy trade-off [2501.03392].

FedFair$^3$ makes adaptation act on **client selection** rather than only aggregation. Its utility score
\[
U_i= |\kappa|\sqrt{\frac{1}{|\kappa|}\times  \sum \limits_{i \in \kappa} {Loss(i)^2}. (\frac{T}{t_i})^{l(T<t_i)\beta}. \lambda^{l(T>t_i)\gamma_i}
\]
depends on local loss, data size, training time, energy consumption, computational capability, and a penalty for excessive time, with normalized selection probability
\[
p_i=\frac{U_i}{\sum U_i}.
\]
The paper’s algorithm then samples clients according to this priority distribution while monitoring a resource-budget constraint [2401.16350].

The diversity of these mechanisms is summarized below.

| Method | Adaptive mechanism | Fairness target |
|---|---|---|
| FairFed | Round-by-round fairness-gap-based aggregation weights | Group fairness |
| FAIR-FATE | Fairness-aware aggregation with Momentum, $\lambda_t$, $\beta_t$ | Group fairness |
| AdaFed | Adaptive common descent direction from orthogonalized, loss-scaled gradients | Client performance fairness |
| FedFair$^3$ | Utility-aware client selection with penalties and resource monitoring | Selection, participation, and accuracy-distribution fairness |
| FedAA | DDPG-based continuous aggregation weights and distance-based client selection | Performance fairness with robustness |
| FedFDP | Client-specific step-size scaling and fairness-aware clipping under DP | Balanced client fairness with privacy |
| PraFFL | Preference-conditioned hypernetwork for personalized trade-offs | Fairness–performance preference matching |
| AAggFF | Online convex optimization of aggregation weights | Client-level fairness |
| OTA-FFL | Modified Chebyshev round-wise adaptive weighting | Performance fairness |

## 4. Extensions: privacy, robustness, personalization, and distribution learning

FedAA shows how AFFL can be combined with **robustness against malicious clients**. It uses parameter-distance-based client selection inspired by FABA, constructs the RL state from selected row-sum distances, lets the server act as a DDPG agent, and chooses continuous aggregation weights over the selected top-$M\%$ clients. Its reward is validation accuracy on a fair held-out validation set, and the paper explicitly defines robustness under Byzantine attack as higher mean test accuracy across benign clients [2402.05541].

FedFDP couples AFFL with **differential privacy**. Starting from the fairness-aware objective
\[
H(\mathbf{w}) = F(\mathbf{w}) + \frac{\lambda}{2} \sum_{i=1}^{N} \left( F_i(\mathbf{w}) - F(\mathbf{w}) \right)^2,
\]
it obtains a client-specific effective step size $\eta_i=\eta(1+\lambda\Delta_i)$ and then introduces a fairness-aware clipping coefficient
\[
C^{i,j}_t = \min\left( 1+\lambda\cdot\Delta_{i}^j,\frac{C}{\| \nabla F_i(\mathbf{w}_t^i,\xi_j) \|} \right).
\]
It also clips and privatizes uploaded loss values using an adaptive clipping threshold derived from the previous round’s privatized mean loss, with privacy analyzed in Rényi Differential Privacy [2402.16028].

PraFFL addresses a different issue: the assumption that each client has only a single pre-defined fairness–accuracy preference. It uses a weighted Tchebycheff scalarization and a client-side hypernetwork $\bm{\beta}_k$ to map a preference vector $\bm{\lambda}$ directly to the personalized parameters $\phi_k = \bm{\beta}_k(\bm{\lambda})$. This is adaptive at inference time because a client can supply an arbitrary preference and obtain the corresponding personalized model without retraining the full FL system [2404.08973].

AFed shifts the locus of adaptation from aggregation to **distribution learning and synthetic augmentation**. AFed-G trains a conditional generator on the server side in latent space; AFed-GAN trains a conditional GAN on clients and aggregates the generative parameters. The classifier is then trained with a combined accuracy and fairness objective using generated latent samples and a mixup-style fairness regularizer [2501.02732].

OTA-FFL incorporates AFFL into a **wireless multiple-access-channel** setting. The adaptive weights are recomputed each communication round from current client losses, then embedded into over-the-air analog aggregation through optimal transmit scalars and a de-noising scalar chosen to make the aggregated gradient unbiased with minimum MSE [2501.03392].

Two papers broaden the conceptual perimeter of AFFL. The medical-AI framework titled “Beyond Static Knowledge Messengers” proposes AFFL as a vision combining adaptive messenger capacity, fairness-aware distillation, and curriculum-guided acceleration; its convergence and fairness statements, as well as its reported efficiency and equity gains, are explicitly framed as projected or feasibility-study outcomes rather than a completed benchmark study [2510.06259]. By contrast, Adaptive FedECADO is described as **not a fairness paper in the usual AFFL sense**; it is better regarded as AFFL-adjacent because it adapts client and server learning rates and momentum to heterogeneity without introducing a fairness objective [2510.04203].

## 5. Experimental practice and reported results

The evaluation practice in AFFL is heterogeneous because the fairness targets are heterogeneous. Client-performance papers commonly report **average accuracy**, **standard deviation or variance across clients**, and **worst-client or worst-10\% performance**. Group-fairness papers report SP, EO, EQO, EOD, SPD, or DP disparity. Some papers add wall-clock time, hypervolume over trade-off sets, Gini coefficient, or privacy budgets [2401.04993][2405.20821].

FedFair$^3$ reported a headline reduction of **18.15% less accuracy variance on IID data** and **54.78% on non-IID data**, together with **24.36% less wall-clock training time on average**, “without decreasing the global accuracy.” On CIFAR10, the reported variance dropped to **3.25** for FedFair$^3$ vs **3.59** for Oort under IID and to **10.32** vs **13.13** under non-IID; similar reductions were reported for FashionMNIST and MNIST [2401.16350].

FairFed’s main empirical evidence is in group fairness under heterogeneity. For $\alpha = 0.1$, it improved EOD by about **93%** on Adult with only about **0.3% accuracy drop**, and by about **50%** on COMPAS with similarly small accuracy cost. In the ACSIncome case study it improved EOD by about **20%**, and on TILES it improved EOD from about **-0.199 to 0.004** [2110.00857].

FAIR-FATE reports broad fairness gains under heterogeneity: on COMPAS it surpasses baselines in **78% of cases**; on Adult it consistently surpasses baselines with about **10% average fairness improvement**; on Law School and Dutch Census it surpasses baselines in **78%** and **88%** of cases, respectively. For strong heterogeneity with $\sigma=0.5$, it reports gains such as **+7% SP, +11% EO, +16% EQO** on COMPAS and **+8% SP, +7% EO, +19% EQO** on Adult [2209.13678].

FedFDP evaluates the privacy–fairness–utility trade-off under $\epsilon=3.52$. The main table reports **95.13% accuracy, $\Psi = 2.3e10$** on MNIST, **85.99% accuracy, $\Psi = 2.8e8$** on FashionMNIST, and **54.21% accuracy, $\Psi = 2.6e9$** on CIFAR10. It states that, when accuracy matches the best baseline, fairness improves by **30.3%** on MNIST, **6.7%** on FashionMNIST, and **27.8%** on CIFAR10 [2402.16028].

PraFFL evaluates the quality of fairness–performance trade-off sets by the 2D hypervolume indicator. It reports best HV values of about **0.971** on SYNTHETIC, **0.772** on COMPAS, **1.100** on BANK, and **1.010** on Adult, outperforming all baselines in the provided comparison [2404.08973].

AdaFedAdam emphasizes fairness with acceleration. On Femnist, the reported averages are **84.48%** accuracy, **8.62%** STD, and **74.16%** worst 30% for AdaFedAdam, versus **82.97%**, **11.44%**, and **67.65%** for FedAdam, and **77.77%**, **13.20%**, and **60.11%** for FedAvg. On Cifar10, AdaFedAdam reports **62.81%** average accuracy, **8.18%** STD, and **46.01%** worst 30%, improving on the listed baselines [2301.09357]. AdaFed reports similarly strong fairness behavior on FEMNIST, including $\sigma_a=6.58$ on FEMNIST-original and, on FEMNIST-skewed, $\bar a=92.21$, $\sigma_a=7.56$, and $\text{KL}=0.011$ [2401.04993].

FeDABoost reports **24.4% reduction** in client F1 variance on MNIST, with **0.0103** variance versus **0.0137** for FedAvg. On FEMNIST it reports median F1 **0.652** and variance **0.0279**, versus **0.558/0.0296** for FedAvg and **0.566/0.0317** for Ditto [2510.02914]. AAggFF reports better degree of client-level fairness than existing methods in both cross-device and cross-silo settings and supplies sublinear regret guarantees for both regimes [2405.20821]. OTA-FFL reports consistent reductions in client disparity, the lowest or near-lowest standard deviation in most datasets, and notable improvements in worst-10\% accuracy [2501.03392].

## 6. Limitations, assumptions, and unresolved issues

Several limitations recur. Many analyses rely on restrictive optimization assumptions. FedFair$^3$ explicitly assumes **convex** local losses; FedFDP assumes $L$-smoothness, $\mu$-strong convexity, and bounded stochastic gradients; AdaFedAdam’s proofs are for convex $L$-smooth losses with bounded gradients; AdaFed’s theory assumes linearly independent client gradients and standard smoothness conditions [2401.16350][2402.16028][2301.09357][2401.04993]. This suggests that a substantial part of AFFL theory remains idealized relative to practical deep-learning regimes.

A second limitation concerns **information requirements**. FAIR-FATE requires a **server-side validation set with sensitive labels**. FairFed assumes secure aggregation and access to global sensitive-attribute marginals computed securely. AFed relies on sensitive attributes and latent-space generative modeling. FedFDP requires privatized uploaded loss values because the server must compute the global loss needed for fairness adjustment [2209.13678][2110.00857][2501.02732][2402.16028].

A third issue is that “adaptive” itself is used in different strengths. FAIR-FATE is described as only **partially adaptive**, because its adaptation is mainly via predefined schedules for $\lambda_t$ and $\beta_t$ rather than a fully learned or feedback-driven controller [2209.13678]. By contrast, FairFed, FedFair$^3$, AAggFF, FedAA, and OTA-FFL adapt weights or selection decisions from current fairness or utility signals at each round [2110.00857][2401.16350][2405.20821][2402.05541][2501.03392].

Finally, no single fairness definition dominates the field. Performance variance across clients, worst-client accuracy, group fairness metrics, participation frequency, and preference-conditioned personalization are all treated as primary targets in different papers. This suggests that AFFL is not converging toward one canonical fairness objective; instead, it is becoming a framework for embedding different fairness notions into federated optimization under non-IID data, privacy constraints, and system heterogeneity.

Source: https://www.emergentmind.com/topics/adaptive-fair-federated-learning-affl