GerryChain.NET

A .NET implementation of the ReCom Markov chain for redistricting, with parallelized proposal generation. Development goals target fast and flexible heuristic optimization.

TODO::

• Implement flips Proposal version of PCompress.Replayer

• Create example scripts for:

  • Optimization
  • PCompress
  • Displacement
  • VRA Scores

• Scores to make factories for:

  • Tally
  • CutEdges
  • Election Results
  • Partisan Metrics
    • Mean Median
    • Efficiency Gap
    • Eguia's Metric

• Add Proposal Generators

  • ReCom
    • ReCom vanilla (cut edge)
    • ReCom vanilla (district pairs)
    • Reversible ReCom
  • Flip chain
  • Swap chain

• Lazy evaluated score on Partition objects

• Decision on using Quikgraph or "homebrewed" immutable graph representation? - Using Quikgraph

• Add grid graph toy example

• Add unit test for initial graph representation

• Add json Parser for networkx format

• Create ReCom/Chain interface and restructure proposal generation to implement with support for multiple ReCom variants.

• Add Chain class

  • ReCom step function (with parallelized proposal generation)
  • Neutral ensemble sample
  • Add support for County Aware chains. But add region crossing tag to graph nodes that encode the scale of the edge weight.

• Add Optimization chain class

  • Objective function representation.
  • Short Bursts
  • Modified Hill Climbing