Computation Engine

• Monte Carlo Markov Chain (MCMC) iteration
  • General facility for Markov Chain testing and development
  • Provision for extensible convergence testing
  • Easy to add new chain generation algorithms and convergence models (e.g. convergence acceleration schemes such jump diffusion, multiple chains)
• Astronomical data sets require line-of-site model integration
  • Can substitute a variety of quadrature/cubature methods
  • Controllable/adaptive caching/interpolation/recomputation is under active development
• An ``Empirical Bayes'' hierarchical priors approach, leading to progressive refinement of spatial/temporal detail (see Figure 1 and Figure 2.)
  • Uniform or non-uniform progression to finer granularity
  • Extensible control of the refinement
  • Extensible tessellation techniques
• Data spatially associated with tiles (spatial regions) in a hierarchy (see Figure)
  • Data sets processed to distributions in general/extensible manner
  • Supports multiple distributions, and multi-dimensional distributions
  • Can easily add new distribution representations (such as basis sets)
• Persistence subsystem (in design; not yet implemented)
  • Save intermediate/processed data sets, by name and with full account of how they were derived from inputs
  • Maintain and check dependences of data sets on one another
  • Save/restore under user control for checkpointing and what-if scenarios