def get_dragon_kings_abs(cutoff_count=1112550, min_size=200, index=None):
if index is None:
import fileio
index = fileio.get_index_table()
heavy_ids = index[
(index.count_span>cutoff_count) &
(index.n_samples>min_size)
].index.values
return heavy_ids
def get_whatever(N=100, min_size=200, index=None):
import numpy as np
if index is None:
import fileio
index = fileio.get_index_table()
vids = index[
(index.n_samples>min_size)
].index.values
return np.random.choice(vids, N, replace=False)
def get_dragon_kings_rel(cutoff_frac=0.75, index=None):
import numpy as np
if index is None:
import fileio
index = fileio.get_index_table()
n_occ = index.count_span.values
n_occ = np.sort(n_occ)
cum_occ = np.cumsum(n_occ)
cutoff_count = n_occ[np.searchsorted(cum_occ, cum_occ[-1]*cutoff_frac)]
return get_dragon_kings_abs(cutoff_count, index=index)
def get_tricky_cases(min_size=200, index=None, shuffle=False):
import numpy as np
if index is None:
import fileio
index = fileio.get_index_table()
s = set(get_flagships(index=index))
s = s.union(get_lead_balloons(index=index, min_size=min_size, N=300))
s = s.union(get_dragon_kings_abs(index=index, min_size=min_size))
s = np.array(list(s))
if shuffle:
np.random.shuffle(s)
return list(s)
def get_lead_balloons(cutoff_time=0.03,
quantile="c50",
min_size=200,
N=None,
index=None):
import numpy as np
if index is None:
import fileio
index = fileio.get_index_table()
vids = index[
(index[quantile+"_sample"]<cutoff_time) &
(index.n_samples>min_size)
].index.values
if N is None:
return vids
return np.random.choice(vids, N, replace=False)