def mergeSort(arr: list,
comp: Comparator,
statParams: list = None,
n: int = 2,
combGroups: bool = True,
sortGroups: bool = False) -> list:
"""Sorts the array with the given comparator.
statParams must be the format ((D0, D1), dist, target AUC)
combGroups determins if the returned array is one list or each group as its own list.
sortGroups determins if groups will be sorted by size in the sort.
yields the arr after each pass through
also yields the stats if given statParams"""
groups: list = list([arr[i]] for i in range(len(arr)))
mergers: list = []
currLayer: int = -1
# while there are partitions
while len(groups) != 1:
currLayer += 1
i: int = 0
while len(groups) >= n:
# last group, odd one out
# get n arrays
# feed the MultiMergers with them
arrays: list = [groups.pop(0) for _ in range(n)]
mergers.append(MultiMerger(arrays, comp, i, 0))
i += 1
#while we have active mergers
while mergers:
for merger in mergers:
res = merger.inc()
if res: #if that merger is done
validate(merger.output)
comp.learn(merger.output)
if sortGroups:
groups.append(merger.output)
else:
groups.insert(0, merger.output)
mergers.remove(merger)
if combGroups:
arr: list = []
for group in groups:
arr.extend(group)
else:
arr: list = groups
# run dem stats
if statParams:
stats: list = runStats(groups, statParams +
[n, currLayer, len(mergers)], comp)
yield arr, stats
else:
yield arr
def analyzeAFCStudies(log: str, results: str, n0: int, n1: int) -> tuple: """Extracts the times out of the log file generated from DylAFC extracts the x0 and x1 vectors and the ranks from the results file from DylComp""" times = list() with open(log) as f: for line in f: line: list = line.strip().split() times.append(float(line[-1])) data, D0, D1 = continuousScale(n0, n1) comp = Comparator(data, rand=True) comp.learn(results) for arr in treeMergeSort(data[:], comp): pass indeciesAFC: list = [arr.index(i) for i in range(256)] x0, x1 = genX0X1(arr, D1, D0) x0: np.ndarray = np.array([indeciesAFC[i] for i in range(128)]) x1: np.ndarray = np.array([indeciesAFC[i] for i in range(128, 256)]) return times, x0, x1, indeciesAFC
def sort(args) -> list:
"""Performs a sort based on the given args.
Args is of the format (dist, auc, n0, n1) and is one tuple/list.
Throws an error if the array did not sort correctly.
Returns the results."""
dist, auc, n0, n1 = args
results = list()
data, D0, D1 = continuousScale(n0, n1)
comp = Comparator(data, level=0, rand=True)
sep = genSep(dist, auc)
comp.genRand(n0, n1, sep, dist)
for arr, stats in treeMergeSort(data, comp, [(D0, D1), dist, auc], n=2):
stats.extend([len(comp), comp.genSeps(), comp.pc[-1]])
comp.resetPC()
results.append(stats)
if arr != sorted(arr, key=lambda x: comp.getLatentScore(x)[0]):
print(arr)
print(sorted(arr, key=lambda x: comp.getLatentScore(x)[0]))
raise AssertionError("did not sort")
return results
anim = FuncAnimation(fig,
update,
frames=np.arange(0, frames),
interval=100)
anim.save("rocs.gif", writer=PillowWriter(fps=10))
pbar.close()
else:
import matplotlib.pyplot as plt
from apng import APNG
from DylSort import treeMergeSort
from DylComp import Comparator
from DylData import continuousScale
from DylMath import genROC, avROC
seed = 15
data, D0, D1 = continuousScale(128, 128)
comp = Comparator(data, level=0, rand=True, seed=seed)
comp.genRand(len(D0), len(D1), 7.72, 'exponential')
np.random.seed(seed)
im = APNG()
fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2)
fig.suptitle("Pass number, MSE true, MSE empirical")
x = np.linspace(0, 1, num=200)
y = x**(1 / 7.72)
ax1.set_aspect('equal', 'box')
ax2.set_aspect('equal', 'box')
elo = simulation_ELO_targetAUC(True)
merge = treeMergeSort(data,
comp,
statParams=[(D0, D1)],
combGroups=False)
plt.tight_layout()
arrays: list = [[0, 1, 4, 2, 5, 3, 6],
[0, 1, 2, 4, 3, 5, 6],
[0, 1, 2, 4, 3, 5, 6],
[0, 1, 2, 3, 4, 5, 6]]
graphROCs(arrays, D0=[0, 1, 2, 3], D1=[4, 5, 6])
elif test == 5:
graphROC([4, 1, 2, 3], [1, 2], [3, 4])
elif test == 6:
from DylSort import treeMergeSort
from DylComp import Comparator
from DylData import continuousScale
import matplotlib
font: dict = {'size' : 10}
matplotlib.rc('font', **font)
data, D0, D1 = continuousScale(128, 128)
comp: Comparator = Comparator(data, rand=True, level=0, seed=15)
for arr in treeMergeSort(data, comp=comp):
pass
D0.sort(key = comp.getLatentScore)
D1.sort(key = comp.getLatentScore)
roc: dict = ROC1.rocxy(comp.getLatentScore(D1), comp.getLatentScore(D0))
graphROCs([arr], True, True, D0, D1)
elif test == 7:
roc1: list = [[0, 0], [0, 1], [1, 1]]
roc3 = roc2 = roc1
roc4: list = [[0, 0], [0.5, 0], [0.5, 0.5], [1, 1]]
avgROC: tuple = avROC([roc1, roc2, roc3, roc4])
fig = plt.figure(figsize=(4,4))
ax = fig.add_subplot(111)
ax.plot(*zip(*roc1), 'm', label='chunk1', ls='-')
ax.plot(*zip(*roc2), 'b', label='chunk2', ls='--')
for group in groups: arr.extend(group)
else:
arr: list = groups
yield (arr, runStats(groups, statParams + [n, layer, len(mergerss)], comp)) if statParams else arr
if __name__ == "__main__":
test: int = int(argv[1]) if len(argv) > 1 else 1
if test == 1:
if len(argv) > 5 or len(argv) < 4:
print("Usage:")
print(f"{__file__} 1 <n0> <n1> <directory to save file into (optional)>")
else:
import matplotlib.pyplot as plt
plt.rcParams["font.size"]: int = 10
data, D0, D1 = continuousScale(int(argv[2]), int(argv[3]))
comp: Comparator = Comparator(data, rand=True)
for arr in treeMergeSort(data, comp):
pass
arrays: list = [arr]
D0.sort(key=arr.index)
D1.sort(key=arr.index)
plt = graphROCs(arrays, True, D0=D0, D1=D1)
ax = plt.gca()
ax.set_title("")
plt.title("")
plt.gcf().suptitle("")
if len(argv) > 4:
plt.savefig(argv[4] + "/patches.pdf", bbox_inches = 'tight', pad_inches = 0)
else:
plt.show()
elif test == 3:
indeciesRight.pop(i)
if len(groups) == 1:
for i in range(indecies[0], indeciesRight[0] + 1):
output[OIndex] = groups[0][i]
OIndex += 1
yield output
left ^= toggle # toggle left if toggle == True
yield False
if __name__ == '__main__':
from DylComp import Comparator
for test in range(1, 7):
if test == 1:
objs: list = [*range(8)]
comp: Comparator = Comparator(objs, rand=False)
m: MultiMerger = MultiMerger([[0, 4], [1, 5], [2, 6], [3, 7]],
comp,
toggle=True)
while not m.inc():
pass
if m.output != [0, 1, 2, 3, 4, 5, 6, 7]:
raise AssertionError("wasn't right")
comp.clearHistory()
m1: MultiMerger = MultiMerger([[0, 4], [1, 5]], comp, toggle=True)
while not m1.inc():
pass
if m1.output != [0, 1, 4, 5]:
raise AssertionError("wasn't right")
m2: MultiMerger = MultiMerger([[2, 6], [3, 7]], comp, toggle=True)
while not m2.inc():