def load_features(self):
(self.sites, self.site_names, self.features, self.feature_names,
self.state_names, self.states, self.families, self.family_names,
self.log_load_features) = read_features_from_csv(
file=self.config['data']['features'],
feature_states_file=self.config['data']['feature_states'])
self.network = compute_network(self.sites, crs=self.crs)
def read_data(self):
print('Reading input data...')
if self.is_simulation:
self.sites, self.site_names, _ = read_sites(self.path_data)
else:
self.sites, self.site_names, _, _, _, _, self.families, self.family_names, _ = \
read_features_from_csv(self.path_data, self.path_feature_states)
self.network = compute_network(self.sites)
self.locations, self.dist_mat = self.network[
'locations'], self.network['dist_mat']
def load_features(self):
self.features = []
c_id = self.ds["ParameterTable", 'id'].name
for feature in self.ds["ParameterTable"]:
self.features.append(feature[c_id])
self.features = numpy.array([[1, 2]])
c_id = self.ds["LanguageTable", 'id'].name
c_name = self.ds["LanguageTable", 'name'].name
c_lon = self.ds["LanguageTable", 'longitude'].name
c_lat = self.ds["LanguageTable", 'latitude'].name
self.sites = Sites(*zip(*[(site[c_id], (site[c_lon], site[c_lat]),
site[c_name])
for site in self.ds["LanguageTable"]]))
self.network = compute_network(self.sites)
def run_simulation(self):
self.inheritance = self.config['INHERITANCE']
self.subset = self.config['SUBSET']
# Get sites
self.sites, self.site_names, self.log_read_sites = read_sites(file=self.sites_file,
retrieve_family=self.inheritance,
retrieve_subset=self.subset)
self.network = compute_network(self.sites)
#
self.areas = assign_area(area_id=self.config['AREA'], sites_sim=self.sites)
# Simulate families
if self.inheritance:
self.families, self.family_names = assign_family(fam_id=1, sites_sim=self.sites)
else:
self.families = None
# Simulate weights, i.e. the influence of universal pressure, contact and inheritance on each feature
self.weights = simulate_weights(i_universal=self.config['I_UNIVERSAL'],
i_contact=self.config['I_CONTACT'],
i_inheritance=self.config['I_INHERITANCE'],
inheritance=self.inheritance,
n_features=self.config['N_FEATURES'])
attempts = 0
while True:
attempts += 1
# Simulate probabilities for features to be universally preferred,
# passed through contact (and inherited if available)
self.p_universal, self.p_contact, self.p_inheritance \
= simulate_assignment_probabilities(e_universal=self.config['E_UNIVERSAL'],
e_contact=self.config['E_CONTACT'],
e_inheritance=self.config['E_INHERITANCE'],
inheritance=self.inheritance,
n_features=self.config['N_FEATURES'],
p_number_categories=self.config['P_N_CATEGORIES'],
areas=self.areas, families=self.families)
correlated = assess_correlation_probabilities(self.p_universal, self.p_contact, self.p_inheritance,
corr_th=self.corr_th)
if correlated <= self.n_correlated:
break
if attempts > 10000:
attempts = 0
self.corr_th += 0.05
self.n_correlated += 1
print("Correlation threshold for simulation increased to", self.corr_th)
print("Number of allowed correlated features increased to", self.n_correlated)
# Simulate features
self.features, self.states, self.feature_names, self.state_names = \
simulate_features(areas=self.areas,
families=self.families,
p_universal=self.p_universal,
p_contact=self.p_contact,
p_inheritance=self.p_inheritance,
weights=self.weights,
inheritance=self.inheritance)
if self.subset:
# The data is split into two parts: subset and complement
# The subset is used for analysis and the complement to define the prior
counts = counts_from_complement(features=self.features,
subset=self.sites['subset'])
self.prior_universal = {'counts': counts,
'states': self.states}
self.network = compute_network(sites=self.sites, subset=self.sites['subset'])
sub_idx = np.nonzero(self.sites['subset'])[0]
self.areas = self.areas[np.newaxis, 0, sub_idx]
self.features = subset_features(features=self.features, subset=self.sites['subset'])