Skip to content

Commit

Permalink
update pyproject
Browse files Browse the repository at this point in the history
  • Loading branch information
@jgibson517
jgibson517 committed Nov 28, 2023
1 parent 8cdc65c commit d3404e2
Show file tree
Hide file tree
Showing 2 changed files with 58 additions and 62 deletions.
2 changes: 1 addition & 1 deletion pyproject.toml
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
[tool.poetry]
name = "votekit"
version = "1.0.2"
version = "1.1.0"
description = "A Swiss Army Knife for computational social choice research"
authors = ["MGGG <engineering@mggg.org>"]
readme = "README.md"
Expand Down
118 changes: 57 additions & 61 deletions src/votekit/ballot_generator.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -29,11 +29,11 @@ class BallotGenerator:
`candidates`.
`pref_interval_by_bloc`
: dictionary mapping of bloc to preference interval.
: dictionary mapping of bloc to preference interval.
(ex. {bloc: {candidate : interval length}}). Defaults to None.
`bloc_voter_prop`
: dictionary mapping of bloc to voter proportions (ex. {bloc: voter proportion}).
: dictionary mapping of bloc to voter proportions (ex. {bloc: voter proportion}).
Defaults to None.
???+ note
Expand All @@ -52,7 +52,6 @@ def __init__(
pref_interval_by_bloc=None,
bloc_voter_prop=None,
):

self.ballot_length = (
ballot_length if ballot_length is not None else len(candidates)
)
Expand Down Expand Up @@ -89,7 +88,7 @@ def from_params(
bloc_voter_prop (dict): A mapping of the percentage of total voters
per bloc (ex. {bloc: 0.7})
cohesion (dict): Cohension factor for each bloc (ex. {bloc: .9})
alphas (dict): Alpha for the Dirichlet distribution of each bloc
alphas (dict): Alpha for the Dirichlet distribution of each bloc
(ex. {bloc: {bloc: 1, opposing_bloc: 1/2}}).
Raises:
Expand Down Expand Up @@ -148,7 +147,7 @@ def _construct_preference_interval(

if "bloc_voter_prop" not in data:
data["bloc_voter_prop"] = bloc_voter_prop

generator = cls(**data)

if isinstance(generator, (AlternatingCrossover, CambridgeSampler)):
Expand All @@ -158,7 +157,7 @@ def _construct_preference_interval(
# rename blocs to match historical data
if isinstance(generator, CambridgeSampler):
generator._rename_blocs()

return generator

@abstractmethod
Expand Down Expand Up @@ -331,7 +330,7 @@ class ImpartialCulture(BallotSimplex):
"""
Impartial Culture model with an alpha value of 1e10 (should be infinity theoretically).
This model is uniform on all linear rankings.
**Attributes**
Expand All @@ -345,16 +344,16 @@ class ImpartialCulture(BallotSimplex):
: dictionary representing a point in the ballot simplex with candidate as
keys and electoral support as values. Defaults to None.
**Methods**
See `BallotSimplex` object.
???+ note
Point or alpha arguments must be included to initialize. For details see
`BallotSimplex` and `BallotGenerator` object.
Point or alpha arguments must be included to initialize. For details see
`BallotSimplex` and `BallotGenerator` object.
"""

def __init__(self, **data):
Expand Down Expand Up @@ -385,8 +384,8 @@ class ImpartialAnonymousCulture(BallotSimplex):
???+ note
Point or alpha arguments must be included to initialize. For details see
`BallotSimplex` and `BallotGenerator` object.
Point or alpha arguments must be included to initialize. For details see
`BallotSimplex` and `BallotGenerator` object.
"""

def __init__(self, **data):
Expand All @@ -396,8 +395,8 @@ def __init__(self, **data):
class PlackettLuce(BallotGenerator):
"""
Class for generating ballots using a Plackett-Luce model. This model samples without
replacement from a preference interval. Can be initialized with an interval or can be
constructed with the Dirichlet distribution using the `from_params` method in the
replacement from a preference interval. Can be initialized with an interval or can be
constructed with the Dirichlet distribution using the `from_params` method in the
`BallotGenerator` class.
**Attributes**
Expand All @@ -418,7 +417,6 @@ class PlackettLuce(BallotGenerator):
"""

def __init__(self, **data):

# Call the parent class's __init__ method to handle common parameters
super().__init__(**data)

Expand All @@ -428,8 +426,9 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:
# the number of ballots per bloc is determined by Huntington-Hill apportionment
blocs = list(self.bloc_voter_prop.keys())
bloc_props = list(self.bloc_voter_prop.values())
ballots_per_block = dict(zip(blocs, apportion.compute("huntington", bloc_props,
number_of_ballots)))
ballots_per_block = dict(
zip(blocs, apportion.compute("huntington", bloc_props, number_of_ballots))
)

for bloc in self.bloc_voter_prop.keys():
# number of voters in this bloc
Expand All @@ -438,11 +437,12 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:
pref_interval_dict = self.pref_interval_by_bloc[bloc]

# finds candidates with non-zero preference
non_zero_cands = [cand for cand, pref in pref_interval_dict.items() if pref > 0]
non_zero_cands = [
cand for cand, pref in pref_interval_dict.items() if pref > 0
]
# creates the interval of probabilities for candidates supported by this block
cand_support_vec = [pref_interval_dict[cand] for cand in non_zero_cands]


for _ in range(num_ballots):
# generates ranking based on probability distribution of non candidate support
non_zero_ranking = list(
Expand All @@ -460,10 +460,10 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:
zero_cands = set(self.candidates).difference(non_zero_cands)
if len(zero_cands) > 0:
ranking.append(zero_cands)

ballot_pool.append(Ballot(ranking = ranking, weight = Fraction(1,1)))

pp = PreferenceProfile(ballots = ballot_pool)
ballot_pool.append(Ballot(ranking=ranking, weight=Fraction(1, 1)))

pp = PreferenceProfile(ballots=ballot_pool)
pp.condense_ballots()
return pp

Expand All @@ -472,10 +472,10 @@ class BradleyTerry(BallotGenerator):
"""
Class for generating ballots using a Bradley-Terry model. The probability of sampling
the ranking $X>Y>Z$ is $P(X>Y)*P(X>Z)*P(Y>Z)$. These individual probabilities are based
on the preference interval: $P(X>Y) = x/(x+y)$. Can be initialized with an interval
or can be constructed with the Dirichlet distribution using the `from_params` method in the
on the preference interval: $P(X>Y) = x/(x+y)$. Can be initialized with an interval
or can be constructed with the Dirichlet distribution using the `from_params` method in the
`BallotGenerator` class.
**Attributes**
`candidates`
Expand All @@ -494,7 +494,6 @@ class BradleyTerry(BallotGenerator):
"""

def __init__(self, **data):

# Call the parent class's __init__ method to handle common parameters
super().__init__(**data)

Expand Down Expand Up @@ -530,22 +529,24 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:
# the number of ballots per bloc is determined by Huntington-Hill apportionment
blocs = list(self.bloc_voter_prop.keys())
bloc_props = list(self.bloc_voter_prop.values())
ballots_per_block = dict(zip(blocs, apportion.compute("huntington", bloc_props,
number_of_ballots)))
ballots_per_block = dict(
zip(blocs, apportion.compute("huntington", bloc_props, number_of_ballots))
)

for bloc in self.bloc_voter_prop.keys():
num_ballots = ballots_per_block[bloc]

pref_interval_dict = self.pref_interval_by_bloc[bloc]
# compute non-zero pref candidates
non_zero_pref_dict = {cand: prop for cand, prop in pref_interval_dict.items() if prop>0}
non_zero_pref_dict = {
cand: prop for cand, prop in pref_interval_dict.items() if prop > 0
}
non_zero_cands = non_zero_pref_dict.keys()
zero_cands = set(self.candidates).difference(non_zero_cands)

# all possible rankings of non zero candidates
permutations = list(it.permutations(non_zero_cands, len(non_zero_cands)))
permutations = list(it.permutations(non_zero_cands, len(non_zero_cands)))


# compute the prob of each ranking given bloc support
ranking_to_prob = self._calc_prob(
permutations=permutations, cand_support_dict=non_zero_pref_dict
Expand All @@ -562,26 +563,28 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:

# sample ballots
for _ in range(num_ballots):
index = list(np.random.choice(
indices,
1,
p=prob_distrib,
))[0]
index = list(
np.random.choice(
indices,
1,
p=prob_distrib,
)
)[0]

# convert index to ranking
ranking = [{cand} for cand in rankings[index]]

# add any zero candidates as ties
if len(zero_cands) > 0 :
if len(zero_cands) > 0:
ranking.append(zero_cands)

ballot = Ballot(ranking = ranking, weight =Fraction(1,1))
ballot = Ballot(ranking=ranking, weight=Fraction(1, 1))
ballot_pool.append(ballot)

# pp = self.ballot_pool_to_profile(
# ballot_pool=ballot_pool, candidates=self.candidates
# )
pp = PreferenceProfile(ballots = ballot_pool)
pp = PreferenceProfile(ballots=ballot_pool)
pp.condense_ballots()
return pp

Expand All @@ -593,9 +596,9 @@ class AlternatingCrossover(BallotGenerator):
or the voters "crossover" and rank a candidate from the other bloc first, then alternate
between candidates from their own bloc and the opposing.
Should only be used when there are two blocs.
Can be initialized with an interval or can be
constructed with the Dirichlet distribution using the `from_params` method in the
Can be initialized with an interval or can be
constructed with the Dirichlet distribution using the `from_params` method in the
`BallotGenerator` class.
**Attributes**
Expand All @@ -612,7 +615,7 @@ class AlternatingCrossover(BallotGenerator):
: dictionary mapping of slate to candidates (ex. {bloc: [candidate1, candidate2]}).
`cohesion_parameters`
: dictionary mapping of bloc to cohesion parameter. A parameter of .6 means voters vote
: dictionary mapping of bloc to cohesion parameter. A parameter of .6 means voters vote
in bloc 60% of time (ex. {bloc: .6}).
**Methods**
Expand All @@ -626,7 +629,6 @@ def __init__(
cohesion_parameters: dict = {},
**data,
):

# Call the parent class's __init__ method to handle common parameters
super().__init__(**data)

Expand All @@ -640,7 +642,6 @@ def __init__(
)

def generate_profile(self, number_of_ballots) -> PreferenceProfile:

ballot_pool = []

# compute the number of bloc and crossover voters in each bloc using Huntington Hill
Expand All @@ -663,7 +664,6 @@ def generate_profile(self, number_of_ballots) -> PreferenceProfile:
)

for bloc in self.bloc_voter_prop.keys():

num_bloc_ballots = ballots_per_type[(bloc, "bloc")]
num_cross_ballots = ballots_per_type[(bloc, "cross")]

Expand Down Expand Up @@ -759,7 +759,7 @@ class CambridgeSampler(BallotGenerator):
blocs, a majority and a minority bloc, and determines this based on the bloc_voter_prop attr.
Based on cohesion parameters, decides if a voter casts their top choice within their bloc
or in the opposing bloc. Then uses historical data; given their first choice, choose a
or in the opposing bloc. Then uses historical data; given their first choice, choose a
ballot type from the historical distribution.
Expand All @@ -769,14 +769,14 @@ class CambridgeSampler(BallotGenerator):
: dictionary mapping of slate to candidates (ex. {bloc: [candidate]}).
`bloc_voter_prop`
: dictionary mapping of bloc to voter proportions (ex. {bloc: voter proportion}).
: dictionary mapping of bloc to voter proportions (ex. {bloc: voter proportion}).
Defaults to None.
`cohesion_parameters`
: dictionary mapping of slate to cohesion level (ex. {bloc: .7}).
`pref_interval_by_bloc`
: dictionary mapping of bloc to preference interval
: dictionary mapping of bloc to preference interval
(ex. {bloc: {candidate : interval length}}).
`historical_majority`
Expand All @@ -802,7 +802,6 @@ def __init__(
historical_minority: Optional[str] = "C",
**data,
):

# Call the parent class's __init__ method to handle common parameters
super().__init__(**data)

Expand All @@ -813,7 +812,7 @@ def __init__(

# changing names to match historical data, if statement handles generating from_params
# only want to run this now if generating from init
if len(self.cohesion_parameters) >0:
if len(self.cohesion_parameters) > 0:
self._rename_blocs()

if path:
Expand All @@ -835,8 +834,10 @@ def _rename_blocs(self):
bloc for bloc in self.bloc_voter_prop.keys() if bloc != majority_bloc
][0]

cambridge_names = {majority_bloc: self.historical_majority,
minority_bloc: self.historical_minority}
cambridge_names = {
majority_bloc: self.historical_majority,
minority_bloc: self.historical_minority,
}

self.slate_to_candidates = {
cambridge_names[b]: self.slate_to_candidates[b]
Expand All @@ -852,26 +853,21 @@ def _rename_blocs(self):
cambridge_names[b]: self.pref_interval_by_bloc[b]
for b in self.pref_interval_by_bloc.keys()
}

self.cohesion_parameters = {
cambridge_names[b]: self.cohesion_parameters[b]
for b in self.cohesion_parameters.keys()
}



def generate_profile(self, number_of_ballots: int) -> PreferenceProfile:

with open(self.path, "rb") as pickle_file:
ballot_frequencies = pickle.load(pickle_file)

ballot_pool = []

# compute the number of bloc and crossover voters in each bloc using Huntington Hill
voter_types = [
(b, t)
for b in list(self.bloc_voter_prop.keys())
for t in ["bloc", "cross"]
(b, t) for b in list(self.bloc_voter_prop.keys()) for t in ["bloc", "cross"]
]

voter_props = [
Expand All @@ -889,7 +885,7 @@ def generate_profile(self, number_of_ballots: int) -> PreferenceProfile:
)

blocs = self.slate_to_candidates.keys()

for bloc in blocs:
# store the opposition bloc
opp_bloc = next(iter(set(blocs).difference(set(bloc))))
Expand Down

0 comments on commit d3404e2

Please sign in to comment.