Inhalt

•  
   List of Figures
•  
   List of Tables
•  
   Glossary
•  
  Introduction
•  
   Peak Intensity Prediction
•  
   Bias and Variance and the Trade-Off
•  
   Why apply Vector-Quantization Based Self-Organizing Maps?
•  
   Why Ensemble Learning?
•  
   Potential of SOM Ensemble Learning
•  
   Outline
•  
   Publications
•  
   Peak Intensity Prediction in a Single Learner Setup
•  
   MS data
•  
   Benchmark Datasets
•  
  Towards Peak Intensity Prediction with Machine Learning methods
•  
   Local Linear Map (LLM) - VQ-based approach
•  
   ν-Support Vector Machine
•  
   Evaluation
•  
   Issues in Data Handling
•  
  Results
•  
   Peptide Prototyping
•  
   Predicting Peaks' Intensities
•  
   Comparison Prediction Performance of SOM to NG
•  
   Subsampling of Peptides
•  
   Comparison Prediction Performance of Feature Sets
•  
   Discussion
•  
   Conclusions
•  
   Improved Peak Intensity Prediction by Adaptive Feature Weighting
•  
   Assessing the Feature Relevance
•  
   Linear Least Squares
•  
   Partial Least Squares
•  
   Random Forests
•  
   Bagged Trees
•  
   Evaluation
•  
   Results
•  
   Weighted Feature Space
•  
   Weighted and Filtered Feature Space
•  
   Discussion
•  
   Conclusions
•  
   Contribution to OpenMS - An Open-Source Framework for MS
•  
   Ensemble Learning
•  
   Reasons for the Success of Ensembles
•  
   Weak Learners
•  
   Unstable Learners
•  
   History of Ensembles
•  
  Assessing Ensemble Error
•  
   Bias-Variance-Covariance Decomposition
•  
   Ambiguity Decomposition
•  
   Assessing Diversity
•  
  Strengthening the Effect by Diverse Predictors
•  
   Bagging
•  
   Random Subspace Method
•  
   Random Forests
•  
   Negative Correlation Learning
•  
   Parameterizing Penalty Functions
•  
  LERRANCO Architecture
•  
   Related Work on SOM Ensemble Learning
•  
   Proposed LERRANCO Architecture
•  
   Accurate and Diverse Ensemble Predictors
•  
   Quantification of Intra-SOM Diversity
•  
   Related Work on SOM Ensemble Learning with NCL
•  
   Conclusion
•  
   SOMs as Accurate and Diverse Ensemble Predictors
•  
   Evaluation
•  
   Training Algorithm
•  
   Topology of Networks
•  
   Training Data
•  
   Size of Random Subspaces
•  
  Initial Conditions
•  
   Grid Size
•  
   Gaussian Neighborhood Width
•  
   Discussion
•  
   Conclusion
•  
  Negatively Correlated SOM Ensembles
•  
   LERRANCO Evaluation
•  
   Results
•  
   Inter-SOM Diversity
•  
   Intra-SOM Diversity
•  
  Dynamics of Boosted Negatively Correlated SOMs
•  
   Dynamics in Time
•  
   Dynamics in λ
•  
   Supporting Altered Penalty Functions
•  
   Discussion
•  
   Complexity and Computation Time
•  
   Conclusion
•  
  Aggregation of Hypotheses to Ensemble Prediction
•  
   Aggregation of Local Experts
•  
   Aggregation of Ensemble Predictors
•  
   Discussion
•  
   Conclusion
•  
   Feature Relevance
•  
   Feature Relevance in the Context of Ensemble Learning
•  
   Random Weighted Subspace Method
•  
   Assessing the Feature Relevance based on OOB a posteriori
•  
   Assessing the Feature Relevance based on Linear Mappings a posteriori
•  
   Robustness of Feature Relevance
•  
   Contribution of Features to Visualization
•  
   Discussion
•  
   Conclusion
•  
  Conclusions
•  
   How SOM ensembles succeed
•  
   Why SOM ensembles succeed
•  
   Outlook
•  
   Summary
•  
   Bibliography