def MSE_compute_1_5(model, n, ITER):
res1 = np.zeros(ITER)
res2 = np.zeros(ITER)
for j in range(ITER):
random.seed(j)
if model == 1:
cov = [[1, 0.9], [0.9, 1]]
beta = 0.9
p_con, p_dis = 0.5, 0.5
gt = -p_con * 0.5 * log(np.linalg.det(cov)) + p_dis * (
log(2) + beta * log(beta) + (1 - beta) *
log(1 - beta)) - p_con * log(p_con) - p_dis * log(p_dis)
x_con, y_con = nr.multivariate_normal([0, 0], cov,
int(n * p_con)).T
x_dis = nr.binomial(1, 0.5, int(n * p_dis))
y_dis = (x_dis + nr.binomial(1, 1 - beta, int(n * p_dis))) % 2
x_dis, y_dis = 2 * x_dis - np.ones(int(
n * p_dis)), 2 * y_dis - np.ones(int(n * p_dis))
x = np.concatenate((x_con, x_dis)).reshape((n, 1))
y = np.concatenate((y_con, y_dis)).reshape((n, 1))
if model == 2:
m = 5
gt = log(m) - 0.8 * log(2)
x = nr.randint(m, size=n)
z = nr.uniform(0, 2, size=n)
y = x + z
if model == 3:
m = 5
gt = 5 * (log(m) - 0.8 * log(2))
x = nr.randint(m, size=5 * n)
z = nr.uniform(0, 2, size=5 * n)
y = x + z
x = x.reshape((n, 5))
y = y.reshape((n, 5))
if model == 4:
p = 0.15
gt = (1 - p) * 0.3012
x = nr.exponential(1, size=n)
z = nr.binomial(1, 1 - p, size=n)
w = nr.poisson(x, size=n)
y = z * w
if model == 5:
p = 0
gt = (1 - p) * 0.3012
x = nr.exponential(1, size=n)
z = nr.binomial(1, 1 - p, size=n)
w = nr.poisson(x, size=n)
y = z * w
res1[j] = (mixed.Mixed_KSG(x, y) - gt) * (mixed.Mixed_KSG(x, y) - gt)
res2[j] = (mixed.KSG(x, y) - gt) * (mixed.KSG(x, y) - gt)
MSE_Mixed_KSG = np.mean(res1)
MSE_copulaksg = np.mean(res2)
return [MSE_Mixed_KSG, MSE_copulaksg]
def estimate_mi(X, y, label, est_H_Y, norm_factor):
if label == "IRF":
frac_eval = 0.3
irf = CalibratedClassifierCV(
base_estimator=RandomForestClassifier(n_estimators=60),
method='isotonic',
cv=5)
# X_train, y_train, X_eval, y_eval = split_train_eval(X, y, frac_eval)
X_train, X_eval, y_train, y_eval = train_test_split(
X, y, test_size=frac_eval)
irf.fit(X_train, y_train)
p = irf.predict_proba(X_eval)
return (est_H_Y - np.mean(entropy(p.T, base=np.exp(1)))) / norm_factor
elif label == "UF":
return (est_H_Y - uf(np.array(X), y)) / norm_factor
elif label == "KSG":
return mixed.KSG(X, y.reshape(-1, 1)) / norm_factor
elif label == "Mixed KSG":
return mixed.Mixed_KSG(X, y.reshape(-1, 1)) / norm_factor
elif label == "UF2":
return (est_H_Y - uf2(np.array(X), y)) / norm_factor
elif label == "UF3":
return UncertaintyForest().fit(X,
y).estimate_mutual_info() / norm_factor
else:
raise ValueError("Unrecognized Label!")
def main(): n = 1000 cov = [[1,0.9],[0.9,1]] beta = 0.9 p_con, p_dis = 0.5, 0.5 gt = -p_con*0.5*log(np.linalg.det(cov))+p_dis*(log(2)+beta*log(beta)+(1-beta)*log(1-beta))-p_con*log(p_con)-p_dis*log(p_dis) x_con,y_con = nr.multivariate_normal([0,0],cov,int(n*p_con)).T x_dis = nr.binomial(1,0.5,int(n*p_dis)) y_dis = (x_dis + nr.binomial(1,1-beta,int(n*p_dis)))%2 x_dis, y_dis = 2*x_dis-np.ones(int(n*p_dis)), 2*y_dis-np.ones(int(n*p_dis)) x = np.concatenate((x_con,x_dis)).reshape((n,1)) y = np.concatenate((y_con,y_dis)).reshape((n,1)) print "Ground Truth = ", gt print "Mixed KSG: I(X:Y) = ", mixed.Mixed_KSG(x,y) print "Partitioning: I(X:Y) = ", mixed.Partitioning(x,y) print "Noisy KSG: I(X:Y) = ", mixed.Noisy_KSG(x,y) print "KSG: I(X:Y) = ", mixed.KSG(x,y)
def main(sample_num, node_num, d, method='KSG'):
mat = np.zeros((trans, all_gene))
value = []
for i in range(trans):
for j in range(all_gene):
if i == j:
continue
if method == 'KSG':
mat[i, j] = mixed.KSG(X[:, i], X[:, j])
elif method == 'Noisy_KSG':
mat[i, j] = mixed.Noisy_KSG(X[:, i], X[:, j])
elif method == 'Partitioning':
mat[i, j] = mixed.Partitioning(X[:, i], X[:, j])
elif method == 'Mixed_KSG':
mat[i, j] = mixed.Mixed_KSG(X[:, i], X[:, j])
elif method == 'mutual_kde':
mat[i, j] = mutual_kde(X[:, i], X[:, j])
value += [mat[i, j]]
#test(method,node_num,mat)
return cal_auc(method, np.array(value))