Baseline Hyperparameter Configurations for Chronic Kidney Disease Prediction Models

Fahman Saeed published a set of baseline hyperparameter configurations on 2026-05-21. The 5.5 KB XLS file likely contains configurations used to train models for predicting Chronic Kidney Disease (CKD) development, specifically for creatinine level prediction from clinical data. The configurations are associated with the Eigen-Guided Transformer model, which was evaluated on the MIMIC-IV and eICU datasets.

Use Cases

• Benchmarking new clinical prediction models based on the provided hyperparameter baselines.
• Reproducing the Eigen-Guided Transformer architecture for creatinine prediction based on the described configuration.
• Analyzing the impact of eigenvector-informed weight initialization on model performance based on the methodology described.
• Studying parameter efficiency in clinical deep learning models based on the reported FLOPs/accuracy ratios.

Strengths

• Specific performance metrics are provided for the associated model, including an MSE of 0.089 and an 11.1% improvement over a baseline.
• The model's fairness was audited, showing a maximum performance discrepancy ratio of 1.32 across demographics.
• The model demonstrated cross-institutional generalizability on the eICU dataset with an MSE of 0.0117.

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 at 5.5 KB, indicating it contains configuration parameters rather than primary clinical data.

Provenance

Source

figshare, author Fahman Saeed

Collection Method

Likely derived from machine learning experiments described in the accompanying research.

Freshness

Last updated 2026-05-21 17:31:24; freshness should be verified.