def test_macau_dense_probit(self):
A = np.random.randn(25, 2)
B = np.random.randn(3, 2)
idx = list(
itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0])))
df = pd.DataFrame(np.asarray(idx), columns=["A", "B"])
df["value"] = (np.array([np.sum(A[i[0], :] * B[i[1], :])
for i in idx]) > 0.0).astype(np.float64)
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
results = macau.macau(Y=Ytrain,
Ytest=Ytest,
side=[A, None],
num_latent=4,
verbose=False,
burnin=20,
nsamples=20,
univariate=False,
precision="probit")
self.assertTrue((results.prediction.columns[0:2] == ["A", "B"]).all())
self.assertTrue(
results.rmse_test > 0.55,
msg=
"Probit factorization (with dense side) gave AUC below 0.55 (%f)."
% results.rmse_test)
def test_macau_tensor_univariate(self):
A = np.random.randn(30, 2)
B = np.random.randn(4, 2)
C = np.random.randn(2, 2)
idx = list(
itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0]),
np.arange(C.shape[0])))
df = pd.DataFrame(np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.array(
[np.sum(A[i[0], :] * B[i[1], :] * C[i[2], :]) for i in idx])
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
Acoo = scipy.sparse.coo_matrix(A)
results = macau.macau(Y=Ytrain,
Ytest=Ytest,
side=[Acoo, None, None],
num_latent=4,
verbose=False,
burnin=20,
nsamples=20,
univariate=True,
precision=50)
self.assertTrue(results.rmse_test < 0.5,
msg="Tensor factorization gave RMSE above 0.5 (%f)." %
results.rmse_test)
def test_bpmf_tensor2(self):
A = np.random.randn(15, 2)
B = np.random.randn(20, 2)
C = np.random.randn(3, 2)
idx = list( itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0]), np.arange(C.shape[0])) )
df = pd.DataFrame( np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.array([ np.sum(A[i[0], :] * B[i[1], :] * C[i[2], :]) for i in idx ])
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
results = macau.bpmf(Y = Ytrain, Ytest = Ytest, num_latent = 4,
verbose = False, burnin = 20, nsamples = 20,
univariate = False, precision = 50)
self.assertTrue(results.rmse_test < 0.5,
msg="Tensor factorization gave RMSE above 0.5 (%f)." % results.rmse_test)
def test_make_train_test_df(self):
idx = list( itertools.product(np.arange(10), np.arange(8), np.arange(3) ))
df = pd.DataFrame( np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.arange(10.0 * 8.0 * 3.0)
Ytr, Yte = macau.make_train_test_df(df, 0.4)
self.assertEqual(Ytr.shape[0], df.shape[0] * 0.6)
self.assertEqual(Yte.shape[0], df.shape[0] * 0.4)
A1 = np.zeros( (10, 8, 3) )
A2 = np.zeros( (10, 8, 3) )
A1[df.A, df.B, df.C] = df.value
A2[Ytr.A, Ytr.B, Ytr.C] = Ytr.value
A2[Yte.A, Yte.B, Yte.C] = Yte.value
self.assertTrue(np.allclose(A1, A2))
def test_macau_dense_probit(self):
A = np.random.randn(25, 2)
B = np.random.randn(3, 2)
idx = list( itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0])) )
df = pd.DataFrame( np.asarray(idx), columns=["A", "B"])
df["value"] = (np.array([ np.sum(A[i[0], :] * B[i[1], :]) for i in idx ]) > 0.0).astype(np.float64)
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
results = macau.macau(Y = Ytrain, Ytest = Ytest, side=[A, None], num_latent = 4,
verbose = False, burnin = 20, nsamples = 20,
univariate = False, precision = "probit")
self.assertTrue( (results.prediction.columns[0:2] == ["A", "B"]).all() )
self.assertTrue(results.auc_test > 0.55,
msg="Probit factorization (with dense side) gave AUC below 0.55 (%f)." % results.rmse_test)
def test_macau_tensor(self):
A = np.random.randn(15, 2)
B = np.random.randn(3, 2)
C = np.random.randn(2, 2)
idx = list( itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0]), np.arange(C.shape[0])) )
df = pd.DataFrame( np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.array([ np.sum(A[i[0], :] * B[i[1], :] * C[i[2], :]) for i in idx ])
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
Acoo = scipy.sparse.coo_matrix(A)
results = macau.macau(Y = Ytrain, Ytest = Ytest, side=[Acoo, None, None], num_latent = 4,
verbose = False, burnin = 20, nsamples = 20,
univariate = False, precision = 50)
self.assertTrue( (results.prediction.columns[0:3] == ["A", "B", "C"]).all() )
self.assertTrue(results.rmse_test < 0.5,
msg="Tensor factorization gave RMSE above 0.5 (%f)." % results.rmse_test)
import unittest
import numpy as np
import pandas as pd
import scipy.sparse
import macau
import itertools
A = np.random.randn(15, 2)
B = np.random.randn(20, 2)
C = np.random.randn(1, 2)
idx = list( itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0]), np.arange(C.shape[0])) )
df = pd.DataFrame( np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.array([ np.sum(A[i[0], :] * B[i[1], :] * C[i[2], :]) for i in idx ])
Ytrain, Ytest = macau.make_train_test_df(df, 0.2)
results = macau.bpmf(Y = Ytrain, Ytest = Ytest, num_latent = 4,
verbose = True, burnin = 20, nsamples = 2,
univariate = False, precision = 50)
Ytrain_sp = scipy.sparse.coo_matrix( (Ytrain.value, (Ytrain.A, Ytrain.B) ) )
Ytest_sp = scipy.sparse.coo_matrix( (Ytest.value, (Ytest.A, Ytest.B) ) )
results = macau.bpmf(Y = Ytrain_sp, Ytest = Ytest_sp, num_latent = 4,
verbose = True, burnin = 20, nsamples = 2,
univariate = False, precision = 50)
def macau_test():
num_latents = 2
n_samples = 800
save_prefix = "macau_unit_test"
## generating toy data
A = np.random.randn(15, num_latents)
B = np.random.randn(3, num_latents)
C = np.random.randn(5, num_latents)
idx = list(
itertools.product(np.arange(A.shape[0]), np.arange(B.shape[0]),
np.arange(C.shape[0])))
df = pd.DataFrame(np.asarray(idx), columns=["A", "B", "C"])
df["value"] = np.array(
[np.sum(A[i[0], :] * B[i[1], :] * C[i[2], :]) for i in idx])
## side information is again a sparse matrix
df_train, df_val = macau.make_train_test_df(df, 0.05)
results = macau.macau(Y=df_train,
Ytest=df_val,
side=[None, None, None],
num_latent=num_latents,
verbose=True,
burnin=400,
nsamples=n_samples,
precision="adaptive",
save_prefix=save_prefix)
str_dir = "results_unittests/"
if (not os.path.exists(str_dir)):
os.makedirs(str_dir)
else:
#replace_prev=input("This configuration has already been run !Do you want to continue ? y/n")
#if (replace_prev=="n"):
# raise ValueError("Aborted")
shutil.rmtree(str_dir)
os.makedirs(str_dir)
files = os.listdir("./")
for f in files:
if (f.startswith(save_prefix)):
shutil.move(f, str_dir)
file_path = save_prefix
N = n_samples
mean_lat_pat = 0
mean_lat_meas = 0
mean_lat_time = 0
for n in range(1, N + 1):
mean_lat_pat += np.loadtxt(str_dir + file_path +
"-sample%d-U1-latents.csv" % n,
delimiter=",")
mean_lat_meas += np.loadtxt(str_dir + file_path +
"-sample%d-U2-latents.csv" % n,
delimiter=",")
mean_lat_time += np.loadtxt(str_dir + file_path +
"-sample%d-U3-latents.csv" % n,
delimiter=",")
mean_lat_pat /= N
mean_lat_meas /= N
mean_lat_time /= N
np.save(str_dir + "mean_pat_latent.npy", mean_lat_pat)
np.save(str_dir + "mean_pat_latent.npy", mean_lat_meas)
np.save(str_dir + "mean_pat_latent.npy", mean_lat_time)
print("Loaded")
print("Patients Latents")
print(mean_lat_pat.T)
print("True Latents")
print(A)
print("Features Latents")
print(mean_lat_meas.T)
print("True Latents")
print(B)
print("Time Latents")
print(mean_lat_time.T)
print("True Latents")
print(C)