Table 3 Training and validation loss for Siamese network on SNIPS dataset.

From: Siamese-based metric joint learning for intent detection and slot filling using triplet loss optimization

Margin

SNIPS dataset

Batch size

16

32

64

128

256

512

Loss

Val_loss

Loss

Val_loss

Loss

Val_loss

Loss

Val_loss

Loss

Val_loss

Loss

Val_loss

0.1

0.00221

0.00628

0.00200

0.00846

0.00200

0.00702

0.00191

0.00552

0.00227

0.00645

0.00336

0.00449

0.2

0.00477

0.01254

0.00297

0.01215

0.00322

0.01166

0.00309

0.01009

0.00393

0.01025

0.00591

0.00715

0.3

0.00475

0.01875

0.00388

0.01533

0.00428

0.01792

0.00453

0.01316

0.00606

0.01420

0.00913

0.01029

0.4

0.00597

0.01522

0.00495

0.01599

0.00557

0.01910

0.00580

0.01702

0.00822

0.01697

0.01270

0.01127

0.5

0.00680

0.02096

0.00605

0.02143

0.00681

0.02240

0.00755

0.02079

0.00845

0.01958

0.01688

0.01218

0.6

0.00721

0.02708

0.00675

0.02571

0.00819

0.02684

0.00941

0.02312

0.00854

0.02478

0.02137

0.01811

0.7

0.00886

0.02947

0.00777

0.02509

0.00928

0.02515

0.01093

0.02578

0.02428

0.01960

0.02428

0.01960

0.8

0.01032

0.03339

0.00923

0.02839

0.01114

0.03364

0.01325

0.02857

0.01970

0.02727

0.02878

0.02662

0.9

0.01202

0.03280

0.01015

0.02753

0.01273

0.04063

0.01461

0.03074

0.02275

0.03720

0.03650

0.02769

1.0

0.01162

0.03805

0.01176

0.03281

0.01256

0.0405

0.01492

0.03164

0.02595

0.04321

0.04175

0.02889

Back to article page