def do_kernel_ridge(stats, library, param_bound):
"This function transforms the accepted parameter values using local linear regression"
X = sm.add_constant(stats)
Y=library
clf = KernelRidge(alpha=1.0, kernel='rbf', coef0=1)
resul = clf.fit(X, Y)
resul_coef=np.dot(X.transpose(), resul.dual_coef_)
coefficients =resul_coef[1:]
para_reg =Y- stats.dot(coefficients)
para_reg=do_ivlogit_transformation(para_reg, param_bound)
parameter_estimate = np.average(para_reg, axis=0)
HPDR=pymc3.stats.hpd(para_reg)
return parameter_estimate, HPDR
def do_kernel_ridge(stats, library, param_bound):
#print('X:', X.shape)
#print('Y:', Y.shape)
#'rbf'
X = sm.add_constant(stats)
Y=library
clf = KernelRidge(alpha=1.0, kernel='rbf', coef0=1)
resul = clf.fit(X, Y)
resul_coef=np.dot(X.transpose(), resul.dual_coef_)
coefficients =resul_coef[1:]
#mean_conf=confidence_interval_Kridge(logit(library), weights, stats,resul_coef)
para_reg =Y- stats.dot(coefficients)
para_reg=do_ivlogit_transformation(para_reg, param_bound)
#param_SS[:,ii] =Y[:,ii]- inv_logit(res_wls_SS.params[1:])+inv_logit(res_wls_OS.params[1:])
parameter_estimate = np.average(para_reg, axis=0)
HPDR=pymc3.stats.hpd(para_reg)
return parameter_estimate, HPDR
