def candidate_regrets(voters, candidates, weights=None, order=1):
"""Calculate the voter regret for each candidate or winner.
Parameters
----------
voters : array (a, n)
Voter preferences; n-dimensional voter cardinal preferences for n issues.
candidates : array (b, n)
Candidate preferences for `b` candidates and `n`-dimensional issues.
Returns
-------
out : (b,) array
Average preference distance of voters from each candidate numbering `b`.
"""
voters = np.atleast_2d(voters)
candidates = np.atleast_2d(candidates)
num_voters = len(voters)
# distance shape (a, b) for `a` num voters, `b` num candidates.
distances = vcalcs.voter_distances(voters,
candidates,
weights=weights,
order=order)
avg_distances = np.sum(distances, axis=0) / num_voters
return avg_distances
def set_raw(self,
voters=None,
weights=None,
order=1,
candidates=None,
winners=None,
distances=None,
ballots=None,
ties=None):
vstat = VoterStats(pref=voters, weights=weights, order=order)
vdata = VoterData(pref=voters,
weights=weights,
order=order,
stats=vstat,
tol=None,
base='linear')
distances = vcalcs.voter_distances(voters=voters,
candidates=candidates,
weights=weights,
order=order)
cstats = CandidateStats(pref=candidates, distances=distances)
cdata = CandidateData(pref=candidates,
distances=distances,
stats=cstats)
edata = ElectionData(ballots=ballots,
winners=winners,
ties=ties,
group_index=None)
self.set_data(voters=vdata, candidates=cdata, election=edata)
return
def regret_std(voters, meanvoter=None, weights=None):
if meanvoter is None:
v_mean = np.mean(voters, axis=0)
else:
v_mean = meanvoter
v_dist = vcalcs.voter_distances(voters, v_mean[None, :], weights=weights)
std = np.std(v_dist)
return std
def nearest_winner_distances(self):
"""array shaped (a,)
Preference distances of nearest winner for `a` voters.
"""
distances = vcalcs.voter_distances(self.voters, self.winners,
self.weights)
index = np.arange(self._num_voters)
return distances[index, self.nearest_winners]
def test2(self):
voters = [[-1, 0, 2]]*5
candidates = [[0,0,0]]*10
voters = np.array(voters)
candidates = np.array(candidates)
d = vcalcs.voter_distances(voters, candidates)
print(d)
self.assertTrue(np.all(d==3))
def _regret_std(voters, meanvoter, weights=None):
#v_mean = np.mean(voters, axis=0)
v_mean = meanvoter
v_dist = vcalcs.voter_distances(voters,
v_mean[None, :],
weights=weights)
std = np.std(v_dist)
return std
def test1(self):
voters = [1,2,3]
candidates = [1,2,3]
voters = np.array(voters).T
candidates = np.array(candidates).T
d = vcalcs.voter_distances(voters, candidates)
print(d)
correct = [[0, 1, 2],
[1, 0, 1],
[2, 1, 0],]
correct = np.array(correct)
self.assertTrue(np.all(d==correct))
return
def calculate_distances(self, candidates):
"""Preference distances of candidates from voters.
Parameters
----------
candidates : array shaped (a, b)
Candidate preference data
"""
pref = self.pref
try:
weights = self.weights
except AttributeError:
weights = None
distances = vcalcs.voter_distances(voters=pref,
candidates=candidates,
weights=weights,
order=self.order)
return distances
def set(self,
voters=None,
weights=-1,
candidates=None,
winners=None,
ballots=None):
"""Set voters, weights, candidates or winners to recalculate"""
if voters is not None:
self._voters = voters
self._cache_voters = {}
self._cache_candidate = {}
self._cache_result = {}
if weights != -1:
self._weights = weights
self._cache_voters = {}
self._cache_candidate = {}
self._cache_result = {}
if candidates is not None:
self._candidates = candidates
self._cache_candidate = {}
self._cache_result = {}
if winners is not None:
self._winners = winners
self._cache_result = {}
if ballots is not None:
self._ballots = ballots
self._distances = vcalcs.voter_distances(voters=self._voters,
candidates=self._candidates,
weights=self._weights,
order=1
)
def calculate_distances(self, candidates):
"""Calculate regret distances.
Parameters
----------
candidates : array shaped (a, b)
Candidate preference data
"""
pref = self._pref
error = self.voter_error
rs = self._randomstate
try:
weights = self.weights
except AttributeError:
weights = None
distances = vcalcs.voter_distances(voters=pref,
candidates=candidates,
weights=weights)
distances = vcalcs.voter_distance_error(distances, error, rstate=rs)
return distances
