Differences

This shows you the differences between two versions of the page.

Link to this comparison view

wisar:tbmod [2014/08/11 19:45] (current)
tmburdge created
Line 1: Line 1:
 +== Terrain-based Bayesian Model ==
  
 +This module is a Bayesian model that uses terrain features and past human behavior data to predict the probability distribution of the likely places to find a missing person. ​
 +
 +== Input/​Output ==
 +Internal Parameters:
 +* Candidate distributions means and standard deviations
 +* Burn (e.g. 1000)
 +* Number of iterations (e.g. 100000)
 +* Slope discretization theshold (e.g. s = 20 degrees)
 +
 +Input Parameters: ​
 +* Search region (defined by GPS coordinates)
 +* Terrain features:
 +** Topography (lake, hill, plane, etc.)
 +** Vegetation density (Sparse, medium, dense)
 +** Elevation (Uphill, flat, downhill)
 +** Trail (On trail, off trail)
 +* Last point seen (GPS coordinates converting to hex coordinates)
 +* Intended destination (GPS coordinates converting to hex coordinates)
 +* Prior Beliefs (transitional probabilities <-- Get from ParamMod)
 +* Desired duration (time since missing person last seen in minute)
 +
 +Output Parameters:
 +* Probability distribution maps at each minute (Array of matrices)
 +
 +== Functions to be implemented ==
 +* Use CUDA to speed up matrix multiplications.
 +** Play with CUDA Net Sample Code
 +** CUDA Test App
 +* Interface to enter search region, last point seen, and (optionally) intended desitnation
 +* Tool to download/​encode topography info (create look up table)
 +* Tool to label topography info (paint tool)
 +* Tool to download/​encode vegetation density (create look up table)
 +* Tool to label trail info (paint tool)
 +* Allow user to specify last point seen and intended destination by clicking a map.
 +* Convert all GPS coordinates to hex coordinates.
 +* Extend model to support trail nodes
 +* Extend model to support intended destination
 +
 +== Validations to be implemented ==
 +* Validate model correctness
 +** Use n-fold cross validation with geocacher GPS track log data
 +** Measure probability density for area within radius r of geocacher position
 +* Validate accuracy of representation
 +** Use past real cases and have both expert and model generate probability distribution maps.
 +** Ask expert to compare maps and reason the differences.
 +** Identify significant differences (exceeding 50% for a certain region)
 +* Validate model usefulness
 +** Use Charles Twardy'​s web site to evaluate models (prior and posterior).
 +** Measure probability density for area within radius r of location where missing person was found.
 +** Measure for multiple r values for sensitivity analysis.
 +** Average densities across all cases to get general sense of how well the model is performing.
 +
 +== Current To Do List ==
 +* Play with CUDA Sample Code
 +* Create CUDA Test App
 +* Implement CUDA in my app
wisar/tbmod.txt ยท Last modified: 2014/08/11 19:45 by tmburdge
Back to top
CC Attribution-Share Alike 4.0 International
chimeric.de = chi`s home Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0