Test Results (20250221_173021)
==================================================

Train size:  300
Val size:  320
Test size:  980
Trainset label distribution:  Counter({3: 40, 5: 35, 1: 32, 4: 31, 0: 30, 2: 29, 7: 28, 6: 27, 9: 26, 8: 22})
Valset label distribution:  Counter({8: 42, 9: 40, 5: 34, 7: 32, 2: 31, 0: 30, 1: 29, 4: 29, 6: 28, 3: 25})
Testset label distribution:  Counter({6: 105, 4: 100, 7: 100, 2: 100, 0: 100, 1: 99, 8: 96, 3: 95, 9: 94, 5: 91})
GAT initialized with 3 layers, [8, 8, 1] heads per layer, [34, 64, 64, 10] features per layer
Dropout: 0.1
Add skip connection: True
Bias: True
Log attention weights: True

Overall Accuracy: 0.9337

Per-class Metrics:
------------------------------

Class 0:
Precision: 0.9065
Recall: 0.9700
F1-score: 0.9372

Class 1:
Precision: 0.9474
Recall: 0.9091
F1-score: 0.9278

Class 2:
Precision: 0.9783
Recall: 0.9000
F1-score: 0.9375

Class 3:
Precision: 0.8824
Recall: 0.9474
F1-score: 0.9137

Class 4:
Precision: 0.8899
Recall: 0.9700
F1-score: 0.9282

Class 5:
Precision: 1.0000
Recall: 0.9231
F1-score: 0.9600

Class 6:
Precision: 0.9462
Recall: 0.8381
F1-score: 0.8889

Class 7:
Precision: 0.9074
Recall: 0.9800
F1-score: 0.9423

Class 8:
Precision: 1.0000
Recall: 0.9479
F1-score: 0.9733

Class 9:
Precision: 0.9091
Recall: 0.9574
F1-score: 0.9326
