Title Page
요약
Abstract
Contents
Chapter 1. Introduction 10
Chapter 2. Related Works 15
2.1. Sparse Autoencoder (SAE) 15
2.2. Label and sparse regularization AE (LSRAE) 16
Chapter 3. Parallel guiding sparse autoencoder 18
Chapter 4. Experiment 26
4.1. Datasets and experimental setups 26
4.2. Parameters Optimization on the four datasets 27
4.3. Classification performance comparison on the three classifiers 34
Chapter 5. Conclusion and future work 38
References 39
Table 1. Summary of four public datasets 26
Table 2. Parameters of PGSAE for four datasets 27
Table 3. Average accuracies of three classifiers by eight dimensionality reduction methods 37
Fig. 1. The virtual structure of AE using ELM 16
Fig. 2. The architecture of PGSAE. It consists of first guiding layer for the reconstruction of guided features and second guiding layer to reflect the distribution... 18
Fig. 3. The classification accuracies on Pima dataset, when ε={0.005 x e|e=1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, λ={0.01 x e|e=... 31
Fig. 4. The classification accuracies on Heart dataset, when ε={0.005 x e|e=1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, λ={0.01 x e|e=... 33
Fig. 5. The classification accuracies on (a) Wine dataset in the best case of ε and (b)Musk dataset in the best case of ε. 34
Fig. 6. The comparison of three different classifiers when 8 dimensionality reduction methods are applied to four datasets. 36