GerryChain.jl

aka

motivation

neato: julia "feels like python, runs like C"

SHAPEFILE_PATH = "./PA_VTD.json"
POPULATION_COL = "TOT_POP"
ASSIGNMENT_COL = "538GOP_PL"

# Initialize graph and partition
graph = BaseGraph(SHAPEFILE_PATH, POPULATION_COL, ASSIGNMENT_COL)
partition = Partition(graph, ASSIGNMENT_COL)

# Define parameters of chain (number of steps and population constraint)
pop_constraint = PopulationConstraint(graph, POPULATION_COL, 0.02)
num_steps = 10

# Initialize Election of interest
election = Election("SEN10", ["SEN10D", "SEN10R"], graph.num_dists)
# Define election-related metrics and scores
election_metrics = [
    vote_count("count_d", election, "SEN10D"),
    efficiency_gap("efficiency_gap", election, "SEN10D"),
    seats_won("seats_won", election, "SEN10D"),
    mean_median("mean_median", election, "SEN10D")
]
scores = [
        DistrictAggregate("presd", "PRES12D"),
        ElectionTracker(election, election_metrics)
]

# Run the chain
println("Running 10-step ReCom chain...")
chain_data = recom_chain(graph, partition, pop_constraint, num_steps, scores)

# Get values of all scores at the 10th state of the chain 
score_dict = get_scores_at_step(chain_data, 10)
# Get all vote counts for each state of the chain
vote_counts_arr = get_score_values(chain_data, "count_d")

running a chain in 34 lines of code!

where is our progress now?

  • Most of core GerryChain is re-implemented in Julia! 🎊
  • Some new features! ✨
  • Package uploaded to Julia registry! 📦
    • using Pkg; Pkg.add("GerryChain")
 
1

some differences w/ gerrychain...

  • "updaters" are now "scores"
    • Tallies âž¡ DistrictAggregate scores
    • introduce district-level vs. plan-level scores
      • saves time + memory
  • acceptance functions must return a float in [0, 1] representing probability of acceptance rather than True or False
    • forces user to think about what acceptance functions are for, reducing confusion w/ constraints!
  • ReCom is first class in GerryChain.jl
    • Python GerryChain very much built on the notion of "flips"
 
👔
 

new features! ✨

  • built-in plotting 📊
  • saving score values to CSV/json files 💾
  • and of course, speed âš¡

The speed of GerryChain depends a lot on how it is configured

Population Constraint:    Tighter the bounds ➡ Longer times

Acceptance Probability:  Lower probability    ➡ Longer times

Number of updaters:       More calculation      ➡ Longer times

Num steps:                         More steps                ➡ Longer times

 

Python Chain Times

Julia Chain Times

Julia and Python Chains, together

So currently we have a ~47x speedup in Julia over Python (over simple runs)

Up next... Parallelization

Up next... Parallelization

thank you!

leave us a 🌟 on github? 🥺👉👈

https://github.com/mggg/GerryChainJulia

GerryChain.jl

By Matthew Sun

GerryChain.jl

  • 340