Title Page
ABSTRACT
초록
Contents
NOMENCLATURE 13
CHAPTER 1. INTRODUCTION 14
CHAPTER 2. DUAL MULTIRESOLUTION NETWORK 17
2.1. Dual Downsampling Block 17
2.1.1. Global Downsampling Module 18
2.1.2. Local Downsampling Module 19
2.2. Feature Extraction Block 20
2.3. Discrimination Block 21
2.4. Anomaly Detection 22
CHAPTER 3. EXPERIMENTS 23
3.1. Experimental settings 23
3.1.1. Datasets 23
3.1.2. Baseline methods 23
3.1.3. Evaluation metric 25
3.1.4. Implementation details 25
3.2. Experimental results 26
3.2.1. Anomaly Detection Performance Comparison 26
3.2.2. Effect of the Downsampling Module 27
3.2.3. Sensitivity to Hyperparameters 30
CHAPTER 4. CONCLUSIONS 32
BIBLIOGRAPHY 33
Table 1. Sliding window description 26
Table 2. Anomaly detection performance comparison 27
Table 3. Effects of global and local downsampling 28
Figure 1. Data with intense oscillation 15
Figure 2. Overall architecture of DResNet 17
Figure 3. Global downsampling module 19
Figure 4. Local downsampling module 20
Figure 5. Feature extraction block 21
Figure 6. Dataset visualization 29
Figure 7. AUROC performance according to S 31