FedEmoNet: Comparison of XAI Methods for Federated Speech Emotion Recognition

5.5 KB of tabular data comparing SHAP and LIME explainability methods within the FedEmoNet framework. The dataset was authored by Mohammed Tawfik and last updated on May 7, 2026. It supports a federated learning system achieving over 99% accuracy on EmoDB and RAVDESS datasets.

Use Cases

• Benchmarking SHAP and LIME interpretability methods based on the dual analysis mentioned in the description
• Analyzing cross-corpus generalization patterns based on the reported accuracy drops for unseen datasets
• Evaluating feature importance for prosodic indicators like fundamental frequency statistics across languages
• Studying the impact of differential privacy on model explainability and attack resistance

Strengths

• High inter-method agreement for XAI techniques is reported (r = 0.997)
• Cross-corpus consistency for prosodic features is noted (r = 0.94)
• Results are derived from a framework achieving high accuracy (over 99% on primary datasets)

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download
• The dataset is very small (5.5 KB), indicating limited scope

Provenance

Source

figshare

Collection Method

Likely contains results from ablation experiments and explainability analysis of the FedEmoNet model.

Time Range

null

Freshness

Last updated 2026-05-07 17:45:43; freshness should be verified

Geography

null