def test5():
numpy.random.seed(42)
data = numpy.zeros((5000, 2))
idx = numpy.random.choice(data.shape[0],
int(data.shape[0] / 2),
replace=False)
data[idx, 1] = 1
idx0 = data[:, 1] == 0
idx1 = data[:, 1] == 1
data[idx0, 0] = numpy.random.normal(100, 30, numpy.sum(idx0))
data[idx1, 0] = numpy.random.normal(200, 30, numpy.sum(idx1))
print(data)
featureNames = ["Gaussian", "Categorical"]
featureTypes = ["continuous", "discrete"]
# spn = SPN.LearnStructure(data, featureTypes=featureTypes, featureNames=featureNames, row_split_method=Splitting.KmeansRows(), col_split_method=Splitting.IndependenceTest(alpha=0.01),
# spn = SPN.LearnStructure(data, featureTypes=featureTypes,
# featureNames=featureNames, row_split_method=Splitting.KmeansRows(),
# col_split_method=Splitting.RDCTest(),
spn = SPN.LearnStructure(data,
featureTypes=featureTypes,
featureNames=featureNames,
row_split_method=Splitting.KmeansRows(),
col_split_method=Splitting.RDCTestOHEpy(),
min_instances_slice=500,
cluster_first=True)
spn.root.validate()
from mpl_toolkits.mplot3d import Axes3D
from matplotlib.collections import PolyCollection
from matplotlib.colors import colorConverter
import matplotlib.pyplot as plt
import numpy as np
fig = plt.figure()
ax = fig.gca(projection='3d')
cc = lambda arg: colorConverter.to_rgba(arg, alpha=0.6)
xs = np.arange(0, 300, 0.5)
verts = []
zs = [0, 1]
maxys = 0
for z in zs:
testdata = numpy.zeros((len(xs), len(zs)))
testdata[:, 0] = xs
testdata[:, 1] = z
ys = numpy.zeros_like(xs)
ys[:] = numpy.exp(spn.root.eval(testdata))
maxys = max(maxys, numpy.max(ys))
ys[0], ys[-1] = 0, 0
verts.append(list(zip(xs, ys)))
poly = PolyCollection(verts, facecolors=[cc('r'), cc('g')])
poly.set_alpha(0.7)
ax.add_collection3d(poly, zs=zs, zdir='y')
ax.set_xlabel('X')
ax.set_xlim3d(0, 300)
ax.set_ylabel('Y')
ax.set_ylim3d(-1, 1)
ax.set_zlabel('Z')
ax.set_zlim3d(0, maxys)
plt.show()
ll = spn.root.eval(data)
print("Sum LL", numpy.sum(ll))
def test6():
numpy.random.seed(42)
datablocks = []
yd = [0, 1, 2, 3]
xd = [0, 1]
for x in xd:
for y in yd:
block = numpy.zeros((2000, 3))
block[:, 1] = x
block[:, 2] = y
if (x == 1 and y == 0) or (x == 0 and y == 1) or (
x == 1 and y == 2) or (x == 0 and y == 3):
block[:, 0] = numpy.random.normal(200, 30, block.shape[0])
else:
block[:, 0] = numpy.random.normal(100, 30, block.shape[0])
datablocks.append(block)
data = numpy.vstack(datablocks)
print(data.shape)
featureNames = ["Gaussian", "Categorical", "Discrete"]
featureTypes = ["continuous", "categorical", "discrete"]
# spn = SPN.LearnStructure(data, featureTypes=featureTypes, featureNames=featureNames, row_split_method=Splitting.KmeansRows(), col_split_method=Splitting.IndependenceTest(alpha=0.01),
# spn = SPN.LearnStructure(data, featureTypes=featureTypes,
# featureNames=featureNames, row_split_method=Splitting.KmeansRows(),
# col_split_method=Splitting.RDCTest(),
spn = SPN.LearnStructure(data,
featureTypes=featureTypes,
featureNames=featureNames,
row_split_method=Splitting.KmeansRows(),
col_split_method=Splitting.RDCTestOHEpy(),
min_instances_slice=50,
cluster_first=True)
spn.root.validate()
from matplotlib.collections import PolyCollection
from matplotlib.colors import colorConverter
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
gs = gridspec.GridSpec(len(xd), len(yd))
fig = plt.figure(figsize=(8, 8))
xall = numpy.arange(0, 300, 0.5)
i = 0
for x in xd:
for y in yd:
testdata = numpy.zeros((len(xall), 3))
testdata[:, 0] = xall
testdata[:, 1] = x
testdata[:, 2] = y
pbs = numpy.zeros_like(xall)
pbs[:] = numpy.exp(spn.root.eval(testdata))
ax = plt.Subplot(fig, gs[i])
i += 1
ax.set_title('%s %s' % (x, y))
ax.plot(xall, pbs, 'r--')
fig.add_subplot(ax)
plt.show()
ll = spn.root.eval(data)
print("Sum LL", numpy.sum(ll))
def test4():
numpy.random.seed(42)
dsname, data, featureNames, featureTypes, doms = getAdult()
data = data[:, [1, 2, 3, 4]]
featureTypes = [
featureTypes[1], featureTypes[2], featureTypes[3], featureTypes[4]
]
featureNames = [
featureNames[1], featureNames[2], featureNames[3], featureNames[4]
]
doms = [doms[1], doms[2], doms[3], doms[4]]
doctorateVal = numpy.where(doms[1] == "Doctorate")[0][0]
stategovVal = numpy.where(doms[0] == "State-gov")[0][0]
print(featureNames)
print(data[0, :])
print(doctorateVal, stategovVal)
pD = numpy.sum(data[:, 1] == doctorateVal) / data.shape[0]
pSD = numpy.sum(
numpy.logical_and(data[:, 1] == doctorateVal, data[:, 0]
== stategovVal)) / data.shape[0]
pS = numpy.sum(data[:, 0] == stategovVal) / data.shape[0]
print("pD", pD)
print("pSD", pSD)
pS_D = pSD / pD
print("pS|D", pS_D)
# spn = SPN.LearnStructure(data, featureTypes=featureTypes, featureNames=featureNames, row_split_method=Splitting.KmeansRows(), col_split_method=Splitting.IndependenceTest(alpha=0.01),
# spn = SPN.LearnStructure(data, featureTypes=featureTypes,
# featureNames=featureNames, row_split_method=Splitting.KmeansRows(),
# col_split_method=Splitting.RDCTest(),
spn = SPN.LearnStructure(data,
featureTypes=featureTypes,
featureNames=featureNames,
row_split_method=Splitting.KmeansRows(),
col_split_method=Splitting.RDCTestOHEpy(),
min_instances_slice=100,
cluster_first=True)
spn.root.validate()
print("SPN Learned")
margSPN_SD = spn.root.marginalizeOut(
[2, 3, 4, 5, 6, 7, 8, 9, 9, 10, 11, 12, 13])
margSPN_SD.Prune()
print(margSPN_SD)
dataSD = numpy.zeros_like(data[0, :]).reshape(1, data.shape[1])
dataSD[0, 0] = stategovVal
dataSD[0, 1] = doctorateVal
print(dataSD)
spnSD = (numpy.exp(margSPN_SD.eval(dataSD)))
margSPN_D = spn.root.marginalizeOut(
[0, 2, 3, 4, 5, 6, 7, 8, 9, 9, 10, 11, 12, 13])
margSPN_D.Prune()
print(margSPN_D)
dataD = numpy.zeros_like(data[0, :]).reshape(1, data.shape[1])
dataD[0, 1] = doctorateVal
print(dataD)
spnD = (numpy.exp(margSPN_D.eval(dataD)))
print("pD", pD)
print("pS", pS)
print("pSD", pSD)
pS_D = pSD / pD
print("pS_D", pS_D)
print("spn pD", spnD)
print("spn pSD", spnSD)
spnS_D = spnSD / spnD
print("spn pS_D", spnS_D)
print("doctorateVal", doctorateVal)
print("stategovVal", stategovVal)
ll = spn.root.eval(data)
# print("Probs", numpy.exp(ll))
print("Sum LL", numpy.sum(ll))
