def test_pyrdc_vs_Rrdc_one_hot_enc():
numpy2ri.activate()
#
# generate random data from two gaussians
rand_gen = numpy.random.RandomState(1337)
n_instances = 20
X_cols = 10
X_size = (n_instances, X_cols)
X_rand = numpy.zeros(X_size)
for i in range(n_instances):
k = rand_gen.choice(X_cols)
X_rand[i, k] = 1
print(X_rand)
Y_cols = 10
Y_size = (n_instances, Y_cols)
Y_rand = numpy.zeros(Y_size)
for i in range(n_instances):
k = rand_gen.choice(Y_cols)
Y_rand[i, k] = 1
print(Y_rand)
print('X: {}, Y: {}'.format(X_rand.shape, Y_rand.shape))
k_values = [1, 2, 10, 20, 100, 200]
for k in k_values:
print('\nConsidering k: {}'.format(k))
#
# generate random normals through R
rnorm = robjects.r["rnorm"]
rnorm_X = rnorm((X_cols + 1) * k)
rnorm_X_m = numpy.array(rnorm_X).reshape(X_cols + 1, k)
rnorm_Y = rnorm((Y_cols + 1) * k)
rnorm_Y_m = numpy.array(rnorm_Y).reshape(Y_cols + 1, k)
print("\trnorm X: {}, rnorm Y: {}".format(rnorm_X_m.shape,
rnorm_Y_m.shape))
#
# loading the R rdc test test from
Rrdc_rank = robjects.r(RRDC_rank)
Rrdc_rank_value = Rrdc_rank(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (rank) value: {}".format(Rrdc_rank_value))
#
# loading the R rdc test test from
Rrdc_ecdf = robjects.r(RRDC_ecdf)
Rrdc_ecdf_value = Rrdc_ecdf(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (ecdf) value: {}".format(Rrdc_ecdf_value))
pyrdc_value = rdc(X_rand, Y_rand, rnorm_X=rnorm_X_m, rnorm_Y=rnorm_Y_m)
print("\tPython value: {}".format(pyrdc_value))
def test_pyrdc_vs_Rrdc_normal_dependent_data():
numpy2ri.activate()
#
# generate random data from two gaussians
rand_gen = numpy.random.RandomState(1337)
n_instances = 5000
X_cols = 100
X_size = (n_instances, X_cols)
X_loc = 10
X_var = 1.0
X_rand = rand_gen.normal(size=X_size, loc=X_loc, scale=X_var)
Y_cols = 100
Y_size = (n_instances, Y_cols)
Y_loc = -25
Y_var = 0.25
alpha = 3.0
beta = 115
Y_rand = alpha * X_rand + beta # rand_gen.normal(size=Y_size, loc=Y_loc, scale=Y_var)
print('X: {}, Y: {}'.format(X_rand.shape, Y_rand.shape))
k_values = [
1, 2, 10, 20, Y_cols, X_cols, X_cols + Y_cols, (X_cols + Y_cols) * 2
]
for k in k_values:
print('\nConsidering k: {}'.format(k))
#
# generate random normals through R
rnorm = robjects.r["rnorm"]
rnorm_X = rnorm((X_cols + 1) * k)
rnorm_X_m = numpy.array(rnorm_X).reshape(X_cols + 1, k)
rnorm_Y = rnorm((Y_cols + 1) * k)
rnorm_Y_m = numpy.array(rnorm_Y).reshape(Y_cols + 1, k)
print("\trnorm X: {}, rnorm Y: {}".format(rnorm_X_m.shape,
rnorm_Y_m.shape))
#
# loading the R rdc test test from
Rrdc_rank = robjects.r(RRDC_rank)
Rrdc_rank_value = Rrdc_rank(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (rank) value: {}".format(Rrdc_rank_value))
#
# loading the R rdc test test from
Rrdc_ecdf = robjects.r(RRDC_ecdf)
Rrdc_ecdf_value = Rrdc_ecdf(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (ecdf) value: {}".format(Rrdc_ecdf_value))
pyrdc_value = rdc(X_rand, Y_rand, rnorm_X=rnorm_X_m, rnorm_Y=rnorm_Y_m)
print("\tPython value: {}".format(pyrdc_value))
def test_pyrdc_vs_Rrdc_bernoulli_data_monodim():
numpy2ri.activate()
#
# generate random data from two gaussians
rand_gen = numpy.random.RandomState(1337)
n_instances = 20
X_cols = 1
X_size = (n_instances, X_cols)
X_theta = .68
X_rand = rand_gen.binomial(n=1, p=X_theta, size=X_size)
print(','.join(str(c[0]) for c in X_rand))
Y_cols = 1
Y_size = (n_instances, Y_cols)
Y_theta = .22
Y_rand = rand_gen.binomial(n=1, p=Y_theta, size=Y_size)
print(','.join(str(c[0]) for c in Y_rand))
print('X: {}, Y: {}'.format(X_rand.shape, Y_rand.shape))
k_values = [1, 2, 10, 20, 100, 200]
for k in k_values:
print('\nConsidering k: {}'.format(k))
#
# generate random normals through R
rnorm = robjects.r["rnorm"]
rnorm_X = rnorm((X_cols + 1) * k)
rnorm_X_m = numpy.array(rnorm_X).reshape(X_cols + 1, k)
rnorm_Y = rnorm((Y_cols + 1) * k)
rnorm_Y_m = numpy.array(rnorm_Y).reshape(Y_cols + 1, k)
print("\trnorm X: {}, rnorm Y: {}".format(rnorm_X_m.shape,
rnorm_Y_m.shape))
#
# loading the R rdc test test from
Rrdc_rank = robjects.r(RRDC_rank)
Rrdc_rank_value = Rrdc_rank(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (rank) value: {}".format(Rrdc_rank_value))
#
# loading the R rdc test test from
Rrdc_ecdf = robjects.r(RRDC_ecdf)
Rrdc_ecdf_value = Rrdc_ecdf(X_rand, Y_rand, rnorm_X, rnorm_Y, k=k)
print("\tR (ecdf) value: {}".format(Rrdc_ecdf_value))
pyrdc_value = rdc(X_rand, Y_rand, rnorm_X=rnorm_X_m, rnorm_Y=rnorm_Y_m)
print("\tPython value: {}".format(pyrdc_value))
def test_pyrdc_vs_Rrdc_bernoulli_data_monodim():
numpy.random.seed(42)
gen = numpy.random.poisson(5, 1000)
gen2 = numpy.random.poisson(5, 1000)
genSorted = numpy.sort(gen)
genmixture = numpy.hstack((gen, numpy.random.poisson(25, len(gen))))
genmixture2 = numpy.hstack((gen2, numpy.random.poisson(25, len(gen2))))
genmixtureSorted = numpy.sort(genmixture)
negdependency = numpy.transpose(numpy.hstack((numpy.vstack(([0] * len(gen), gen)), numpy.vstack((gen, [0] * len(gen))))))
genmixture3 = numpy.hstack((numpy.random.poisson(2, 1000), numpy.random.poisson(13, 2000)))
negdependencymixture = numpy.transpose(numpy.hstack((numpy.vstack(([0] * len(genmixture3), genmixture3)), numpy.vstack((genmixture3, [0] * len(genmixture3))))))
negdependencymixture = negdependencymixture[~numpy.all(negdependencymixture == 0, axis=1)]
independent = numpy.transpose(numpy.vstack((gen, gen2)))
independentmixture = numpy.transpose(numpy.vstack((genmixture, genmixture2)))
opposite = numpy.transpose(numpy.vstack((genSorted, genSorted[::-1])))
oppositemixture = numpy.transpose(numpy.vstack((genmixtureSorted, genmixtureSorted[::-1])))
cooccur = numpy.transpose(numpy.vstack((gen, gen)))
cooccurmixture = numpy.transpose(numpy.vstack((genmixture, genmixture)))
print()
print("cooccur", rdc(cooccur[:,0], cooccur[:,1]))
print("negdependency", rdc(negdependency[:,0], negdependency[:,1]))
print("negdependencymixture", rdc(negdependencymixture[:,0], negdependencymixture[:,1]) )
print("independent", rdc(independent[:,0], independent[:,1]) )
print("independentmixture", rdc(independentmixture[:,0], independentmixture[:,1]) )
print("opposite", rdc(opposite[:,0], opposite[:,1]) )
print("oppositemixture", rdc(oppositemixture[:,0], oppositemixture[:,1]) )
print("cooccur", rdc(cooccur[:,0], cooccur[:,1]) )
print("cooccurmixture", rdc(cooccurmixture[:,0], cooccurmixture[:,1]) )
print(getIndependentGroups(independentmixture, 0.05, "poisson"))