Go to slide: 1: Neural Network Empirical Solutions for Remote Sensing Problems 2: Recommended Reading 3: Outline 4: Continuous Mapping 5: Mapping: examples 6: ANN - Nonlinear Input to Output Mapping 7: A neuron (unit, PE) from inside 8: Activation Function 9: NN as a universal tool for approximation or continuous mapping 10: NN training 11: NNs vs. Regressions 12: Main properties of NNs: 13: Satellite Forward and Inverse Problems As Nonlinear Mappings 14: Satellite Data Utilization 15: Special Sensor Microwave Imager 16: Evolution of the NN architecture for SSM/I retrievals 17: Error budget (m/s) for different SSM/I wind speed algorithms (€15,000 buoy/SSMI matchups) 18: NN vs. GSW: Areal Coverage 19: Storm in Atlantic 1/5/98 20: Storm in Pacific 3/12/99 21: Storm in Atlantic 2/25/98 22: Validation: Buoys vs. GDAS 23: Columnar Water Vapor 24: Columnar Liquid Water 25: NN Empirical Forward Model for SSM/I 26: Comparison of physically based radiative transfer and empirical NN forward models (€7,000 buoy/SSMI matchups) 27: How to Apply NNs 28: How to Apply NNs 29: Conclusions (1) 30: Conclusions (2)