fig.delaxes(ax6)
plt.tight_layout()
plt.savefig("figure.svg")
comp.xVals = xVals
comp.xLabels = xLabels
print(data)
plots = list()
# give dummy 0 vals for dist and target AUC
for currLayer, (groups, stats) in enumerate(
treeMergeSort(data,
comp, [(D0, D1), 0, 0],
combGroups=False)):
print(groups)
rocs = list()
for group in groups:
rocs.append(genROC(group, D0, D1))
avgROC = avROC(rocs)
xLabels[currLayer] = len(comp)
auc, varEstimate, hanleyMcNeil, estimates = stats
f.write(''.join([str(val) + ',' for val in stats]))
f.write('\n')
aucs[currLayer] = auc
varEstimates[currLayer] = varEstimate
hmnEstimates[currLayer] = np.append(
np.full((layers - len(estimates)), np.nan), estimates)
compLens[currLayer] = len(comp)
for plot in plots:
for line in plot:
try:
line.remove()
except ValueError:
roc, mseTheo, mseEmp, empiricROC = next(elo)
ax1.plot(x, y, linestyle='--', label='true', lw=3)
ax1.plot(empiricROC['x'],
empiricROC['y'],
linestyle=':',
lw=2,
label='empirical')
ax1.plot(roc['x'], roc['y'], label='predicted')
ax1.legend(loc=4)
ax1.set_title(
f"ELO\n{i+1}, {mseTheo[0]*1000:02.3f}E(-3), {mseEmp[0]*1000:02.3f}E(-3)"
)
groups = next(merge)
rocs = []
for group in groups:
rocs.append(genROC(group, D0, D1))
roc = avROC(rocs)
mseTheo, mseEmp, auc = MSE(7.72, 'exponential',
zip(*roc)), MSE(
7.72, 'exponential', zip(*roc),
zip(empiricROC['x'],
empiricROC['y']))
ax2.plot(x, y, linestyle='--', label='true', lw=3)
ax2.plot(empiricROC['x'],
empiricROC['y'],
linestyle=':',
lw=2,
label='empirical')
ax2.plot(*roc, label='predicted')
ax2.legend()
ax2.set_title(
matplotlib.use("QT4Agg")
import matplotlib.pyplot as plt
from DylMath import genROC, avROC, auc
data, D0, D1 = continuousScale(16, 16)
comp: Comparator = Comparator(data, rand=True)
comp.genRand(len(D0), len(D1), 7.72, "exponential")
comps: list = list()
rocs: list = list()
overlapping: bool = argv[2] == "True" if len(argv) == 3 else True
for groups in treeMergeSort(data, comp, combGroups=False):
rocs.append(list())
comps.append(len(comp))
arr: list = []
for group in groups:
arr.extend(group)
rocs[-1].append(genROC(group, D0, D1))
rocs[-1]: tuple = list(zip(*avROC(rocs[-1])))
#rocs[-1].reverse()
#print(comp.kendalltau(arr), MSE(7.72, rocs[-1], comp.empiricROC()))
if not overlapping:
rows: int = int(np.ceil(np.sqrt(len(rocs))))
cols: int = int(np.ceil(len(rocs) / rows))
fig, axes = plt.subplots(rows, cols, sharex=True, sharey=True, num="plots")
fig.suptitle("ROC Curves")
for i,ax in enumerate(axes.flat):
if i >= len(rocs):
continue
ax.set_aspect('equal', 'box')
ax.set(xlabel="FPF", ylabel="TPF")
ax.label_outer()
ax.plot((0,1),(0,1),c='red', linestyle=":")