Table 2 Comparison of various evaluation metrics on raw data and data representations

From: GLARE: discovering hidden patterns in spaceflight transcriptome using representation learning

  

Evaluation metrics

Environment

Data representations

Trustworthiness score ↑

KNN accuracy ↑

Silhouette score ↑

FLT

Raw Data

-

65.96 ± 0.39

0.5107

 

t-SNE

0.942

67.67 ± 0.55

0.3842

 

UMAP

0.943

81.36 ± 0.53

0.3985

 

SAE

0.967

90.66 ± 0.36

0.4959

 

FT-SAE

0.888

96.08 ± 0.33

0.5635

GC

Raw Data

-

62.48 ± 0.78

0.5047

 

t-SNE

0.946

64.04 ± 0.82

0.3743

 

UMAP

0.949

80.14 ± 0.54

0.3791

 

SAE

0.951

93.80 ± 0.56

0.5634

 

FT-SAE

0.870

94.45 ± 0.45

0.5886

  1. FT-SAE shows the highest KNN accuracy (with ±standard deviation via 5-fold cross-validation) and the highest Silhouette score while having a lower trustworthiness score compared to others for both FLT and GC. The trustworthiness score is calculated with the raw data and the transformed data representation; therefore, it is not applicable for the raw data alone and is left blank.
  2. Bold values indicate the best-performing results for each column, corresponding to individual evaluation metrics. Metrics marked with ‘↑’ indicate that higher values correspond to better performance
Back to article page