VitalDB - AnesLLM: a dataset and benchmark based on real intraoperative anaesthetic decisions 1.0

File: <base>/reports/clustering_methods.csv (366 bytes)
strategymethod
knnkNN Euclidean K=50 (baseline)
random_forestRandom Forest, probability out-of-fold (3 folds)
lightgbmLightGBM, probabilities out-of-fold (3 folds)
logisticSoftmax logistic regression, out-of-fold (3 folds)
centroidaffinity = softmax(distance to class centroid)
dwknnkNN weighted by 1/distance
umap_knnUMAP (10 dims) + kNN K=50