Neural Network Empirical Solutions for Remote Sensing Problems

• Neural Network Empirical Solutions for Remote Sensing Problems
  
• Recommended Reading
  
• Outline
  
• Continuous Mapping
  
• Mapping: examples
  
• ANN - Nonlinear Input to Output Mapping
  
• A neuron (unit, PE) from inside
  
• Activation Function
  
• NN as a universal tool for approximation or continuous mapping
  
• NN training
  
• NNs vs. Regressions
  
• Main properties of NNs:
  
• Satellite Forward and Inverse Problems As Nonlinear Mappings
  
• Satellite Data Utilization
  
• Special Sensor Microwave Imager
  
• Evolution of the NN architecture for SSM/I retrievals
  
• Error budget (m/s) for different SSM/I wind speed algorithms (€15,000 buoy/SSMI matchups)
  
• NN vs. GSW: Areal Coverage
  
• Storm in Atlantic 1/5/98
  
• Storm in Pacific 3/12/99
  
• Storm in Atlantic 2/25/98
  
• Validation: Buoys vs. GDAS
  
• Columnar Water Vapor
  
• Columnar Liquid Water
  
• NN Empirical Forward Model for SSM/I
  
• Comparison of physically based radiative transfer and empirical NN forward models (€7,000 buoy/SSMI matchups)
  
• How to Apply NNs
  
• How to Apply NNs
  
• Conclusions (1)
  
• Conclusions (2)

---