model = GPy.models.GPRegression(X_train, Y_train, kernel=k)
model.optimize(messages=True, max_iters=100)
# Get predictions
info_dict = {}
if args.model == 'ridge' or args.model == 'svr':
preds = model.predict(X_test)
if args.label_preproc == 'scale':
preds = Y_scaler.inverse_transform(preds)
info_dict['mae'] = MAE(preds, Y_test.flatten())
info_dict['rmse'] = np.sqrt(MSE(preds, Y_test.flatten()))
info_dict['pearsonr'] = pearsonr(preds, Y_test.flatten())
else:
# TODO: check if this makes sense
#preds, vars = model.predict(X_test)
preds, vars = model.predict_noiseless(X_test)
if args.label_preproc == 'scale':
preds = Y_scaler.inverse_transform(preds)
info_dict['mae'] = MAE(preds, Y_test)
info_dict['rmse'] = np.sqrt(MSE(preds, Y_test))
info_dict['pearsonr'] = pearsonr(preds.flatten(), Y_test.flatten())
nlpd = model.log_predictive_density(X_test, Y_test)
info_dict['nlpd'] = np.mean(nlpd)
# Get parameters
if args.model == 'ridge':
info_dict['coefs'] = list(model.coef_)
info_dict['intercept'] = model.intercept_
info_dict['regularization'] = model.alpha_
elif args.model == 'svr':
info_dict['regularization'] = model.best_params_['C']
model = GPy.models.GPRegression(X_train, Y_train, kernel=k)
model.optimize(messages=True, max_iters=100)
# Get predictions
info_dict = {}
if args.model == 'ridge' or args.model == 'svr':
preds = model.predict(X_test)
if args.label_preproc == 'scale':
preds = Y_scaler.inverse_transform(preds)
info_dict['mae'] = MAE(preds, Y_test.flatten())
info_dict['rmse'] = np.sqrt(MSE(preds, Y_test.flatten()))
info_dict['pearsonr'] = pearsonr(preds, Y_test.flatten())
else:
# TODO: check if this makes sense
#preds, vars = model.predict(X_test)
preds, vars = model.predict_noiseless(X_test)
if args.label_preproc == 'scale':
preds = Y_scaler.inverse_transform(preds)
info_dict['mae'] = MAE(preds, Y_test)
info_dict['rmse'] = np.sqrt(MSE(preds, Y_test))
info_dict['pearsonr'] = pearsonr(preds.flatten(), Y_test.flatten())
nlpd = model.log_predictive_density(X_test, Y_test)
info_dict['nlpd'] = np.mean(nlpd)
# Get parameters
if args.model == 'ridge':
info_dict['coefs'] = list(model.coef_)
info_dict['intercept'] = model.intercept_
info_dict['regularization'] = model.alpha_
elif args.model == 'svr':
info_dict['regularization'] = model.best_params_['C']
