def get_loss_info(self, tst_eval_info, val_eval_info):
loss_mN = val_eval_info.get_loss()
w_km = base.build_bootstrap_matrix(self.k_bootstrap, loss_mN.shape[0], self.essr)
prob_N = base.aB_prob(w_km, loss_mN)
loss_m = tst_eval_info.test_distr(prob_N)
return loss_m
def get_loss_info(self, tst_eval_info, val_eval_info):
loss_mN = val_eval_info.get_loss()
w_km = base.build_bootstrap_matrix(self.k_bootstrap, loss_mN.shape[0],
self.essr)
prob_N = base.aB_prob(w_km, loss_mN)
loss_m = tst_eval_info.test_distr(prob_N)
return loss_m
def get_loss_info(self, tst_eval_info, val_eval_info):
loss_mN = val_eval_info.get_loss()
m, N = loss_mN.shape
w_km = base.build_bootstrap_matrix(self.k_bootstrap, m, self.essr)
ensemble_d, prob_d = greedy.greedy_ensemble(loss_mN, w_km, self.ensemble_size)
prob_N = np.zeros(N)
prob_N[ensemble_d] = prob_d
loss_m = tst_eval_info.test_distr(prob_N)
return loss_m
def get_loss_info(self, tst_eval_info, val_eval_info):
loss_mN = val_eval_info.get_loss()
m, N = loss_mN.shape
w_km = base.build_bootstrap_matrix(self.k_bootstrap, m, self.essr)
ensemble_d, prob_d = greedy.greedy_ensemble(loss_mN, w_km,
self.ensemble_size)
prob_N = np.zeros(N)
prob_N[ensemble_d] = prob_d
loss_m = tst_eval_info.test_distr(prob_N)
return loss_m
