|
|
|
Neural Net Spectral Analysis of Icy Volatiles in the Solar System
|
|
|
| We will train and apply a two-stage neural net algorithm to spectra of satellites (Rhea, Mimas, Dione, Europa, Ganymede, Callisto, Titania, Umbriel, Ariel, Oberon, Miranda, Triton and Charon), Centaurs (Chariklo and others) and Pluto all of which have surface water ice or methane ice. These two ices have IR spectral features that are diagnostic of composition, grain size, surface temperature and crystalline state, but the interpretation typically depends on subtle spectral features that defy many spectral classification schemes. The proven neural net algorithm we will use has a self organizing map as a preprocessor to help robustly extract spectral information. This two-tier neural net has already achieved excellent results in several diverse palentary applications. |
|
|
