Contents

• Introduction
  • Background
  • Organization of this work
• Self-Organizing Map
  • Artificial neural networks
  • Self-Organizing Map (SOM)
    • Data preprocessing
      • Focusing on a subset of data
      • Removing erroneous data
      • Data encoding
      • Scaling
    • Initialization
      • Random initialization
      • Initialization using initial samples
      • Linear initialization
    • Training
      • Update rule
      • Neighborhood function
      • Learning rate
    • Visualization
      • U-matrix
      • Sammon's mapping
      • Component plane representation
    • Validation
  • Applications
• Principal component analysis
  • Principal component analysis
• Models
  • General
  • SOM as a regression model
  • SOM and local model fitting
• Case study: Rautaruukki
  • Problem domain
  • Process data
    • Data preprocessing
    • Training a SOM
    • The U-matrix representation
    • Sammon mapping of the SOM
  • Using SOM as a regression model
    • Predicting quality parameters
    • Sensitivity analysis
  • Using SOM and local model fitting
• Conclusions
• References
• About this document ...