Exploring neural network architectures with automated machine learning approaches / submitted by Julian Burghoff, M.Sc. Wuppertal, 23.09.2023
Inhalt
- Acknowledgments
- Foreword
- Contents
- Introduction
- Basics of machine learning and neural networks
- Machine Learning in general
- Feed-Forward Neural Nets
- Activation functions
- Universal Approximation Theorem
- Lossfunction
- Gradient based Optimization
- Backpropagation
- Overfitting
- Types of layers
- Academic Benchmark Datasets
- Port-Hamiltonian Optimizer
- Meta-Learning-Algorithms
- Numerical results on MorphNet
- Numerical results on ResBuilder method
- Experimental setup
- Hyperparameter settings
- Training types
- Explanations of evaluation figures
- LayerLasso Momentum
- Pre-processing the data
- Used Resources
- Numerical results
- Benchmarks
- Removal positions
- Regularization parameter study
- Image manipulation detection in an industrial context
- Parameter study on long time runs
- Study on MorphNet intensity IM on SmallNORB dataset with small initial architecture
- Outline
- Neural Networks in Survival Analysis
- Conclusion
- List of Figures
- List of Tables
- List of Algorithms & Scripts
- List of Notations
- Bibliography