def training_MAP_UB_DUCHI(train, train_r, K, l):
"""
:param train:
:param train_r:
:param K:
:param l:
:return:
"""
k = train_r.shape[1]
K_inv = ghat.train_kernel_inv(train,K,l)
gamma_tr = ghat.gamma_train(train_r, gamma_r.gamma_MAP_UB_DUCHI, k)
return K_inv, gamma_tr
def training_NDCG_rbf(train_r,K,l,k=10):
"""dont forget kernel
:param train:
:param train_r:
:param l:
:param k:
:return:
"""
n = train_r.shape[0]
K_inv = linalg.inv((K + n * l * np.identity(n)))
gamma_tr = ghat.gamma_train(train_r,gamma_r.gamma_NDCG,k)
return K_inv, gamma_tr
def training_DUCHI(train, train_r, l):
k = train_r.shape[1]
K_inv = ghat.train_kernel_inv(l, kernel.kernel_DUCHI, train)
gamma_tr = ghat.gamma_train(train_r, gamma_r.gamma_DUCHI, k)
return K_inv, gamma_tr