Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Table 1.

Qualitative Comparison between IVA-FL and Existing Adaptive DP Paradigms. Unlike existing methods that prioritize convergence or user preference, IVA-FL is the first to prioritize the preservation of minority class signals based on data semantics.

More »

Table 1 Expand

Fig 1.

The overall framework of IVA-FL.

More »

Fig 1 Expand

Table 2.

Core performance comparison of IVA-FL and baseline methods on two imbalanced datasets.

More »

Table 2 Expand

Fig 2.

Trend of AUC performance with privacy budget (ε).

More »

Fig 2 Expand

Fig 3.

Trend of Recall with Class Imbalance (Fixed ε = 4).

More »

Fig 3 Expand

Table 3.

Financial Risk Assessment Metrics Comparison. KS Statistic measures discrimination capability (Higher is better), and Brier Score measures probability calibration (Lower is better). Data is based on the Credit Card Fraud Detection dataset.

More »

Table 3 Expand

Fig 4.

KS Curves Comparison.

The KS curve measures the maximum difference between the cumulative distributions of positive (fraud) and negative (normal) samples. A higher curve peak indicates better risk discrimination capability. IVA-FL demonstrates a significantly higher KS Statistic (0.742) compared to baselines.

More »

Fig 4 Expand

Fig 5.

Robustness Analysis against Data Heterogeneity (Non-IID).

The experiment evaluates model Recall under varying Dirichlet parameters α. A smaller α (e.g., 0.1) indicates extreme data heterogeneity. While baseline methods degrade significantly in extreme Non-IID settings, IVA-FL demonstrates strong resilience, maintaining high utility.

More »

Fig 5 Expand

Table 4.

IVA-FL core mechanism ablation study results (ε = 4).

More »

Table 4 Expand

Fig 6.

Comprehensive Ablation Study: Model Performance Comparison (ε = 4).

More »

Fig 6 Expand

Fig 7.

Scalability Analysis: Final Model Performance vs. Number of Clients.

More »

Fig 7 Expand

Fig 8.

Scalability Analysis: Convergence Speed vs. Number of Clients.

More »

Fig 8 Expand

Table 5.

Comparison of Computational Overhead and Resource Consumption per Communication Round. The experiment was conducted on the Credit Card Fraud Detection dataset with batch size 64.

More »

Table 5 Expand

Fig 9.

Computation Overhead and Efficiency Analysis.

More »

Fig 9 Expand

Fig 10.

Parameter Sensitivity Analysis.

(a) Impact of Clipping Modulation Factor λ on Recall and AUC. (b) Impact of Smoothing Decay Factor γ on Recall and AUC. The shaded areas represent the standard deviation across 5 independent runs. The results indicate that the model achieves optimal performance around and , exhibiting a stable inverted U-shaped trend.

More »

Fig 10 Expand

Table 6.

Model performance comparison in a simulated SCF feature-heterogeneous scenario.

More »

Table 6 Expand

Fig 11.

Performance in Simulated Heterogeneous SCF Scenario.

More »

Fig 11 Expand

Table 7.

Evolution of Model AUC Performance in a Dynamic Market Environment.

More »

Table 7 Expand

Fig 12.

Adapted Performance Comparison After Concept Drift.

More »

Fig 12 Expand