Table 1 Results of experiments with different optimizer and learning rate on ISIC-2018.

From: Dual-channel compression mapping network with fused attention mechanism for medical image segmentation

 

Accuracy

TPR

Dice

Jaccard

Optimizer (learning rate = 0.001)

 Adam

0.9142

0.8893

0.8609

0.7602

 Adamax

0.9128

0.8879

0.8586

0.7582

 RMSprop

0.9090

0.8843

0.8523

0.7526

 Rprop

0.9137

0.8888

0.8601

0.7595

 SGD

0.8965

0.8721

0.8312

0.7340

Learning rate (optimizer = Adam)

 0.0005

0.9091

0.8848

0.8602

0.7596

 0.001

0.9142

0.8893

0.8609

0.7602

 0.005

0.8829

0.8588

0.8092

0.7146

 0.01

0.8752

0.8513

0.7973

0.7040

 0.05

0.8352

0.8124

0.7356

0.6496

 0.1

0.8334

0.8107

0.7325

0.6469

Back to article page