def __init__(self, n_candidates, GRAD_SIZE, EXP_SIZE, k_initial, k_increase, TB_QUEUE_SIZE=None, TB_WINDOW_SIZE=None, prev_qeury_len=None, *args, **kargs):
super(TD_NSGD_DSP, self).__init__(*args, **kargs)
self.model = LinearModel(n_features = self.n_features,
learning_rate = self.learning_rate,
n_candidates = n_candidates)
self.GRAD_SIZE = GRAD_SIZE
self.EXP_SIZE = EXP_SIZE
self.TB_QUEUE_SIZE = TB_QUEUE_SIZE
self.TB_WINDOW_SIZE = TB_WINDOW_SIZE
self.sample_basis = True
self.clicklist = np.empty([self.GRAD_SIZE,1], dtype=int) #click array
self.grad = np.zeros([self.GRAD_SIZE,self.n_features], dtype=float)
self.gradCol = 0
# DQ tie-break related lists
self.difficult_NDCG =[]
self.difficult_queries =[]
self.difficult_document =[]
self.difficult_time =[]
self.query_id = 0
self.k_initial = k_initial
self.k_increase = k_increase
# Secondary techniques
self.prev_qeury_len = prev_qeury_len
if prev_qeury_len:
self.prev_feat_list = []
def __init__(self, learning_rate, learning_rate_decay, *args, **kargs):
super(TD_DBGD, self).__init__(*args, **kargs)
self.learning_rate = learning_rate
self.model = LinearModel(n_features=self.n_features,
learning_rate=learning_rate,
n_candidates=1,
learning_rate_decay=learning_rate_decay)
self.multileaving = TeamDraftMultileave(n_results=self.n_results)
def __init__(self, learning_rate, learning_rate_decay, *args, **kargs):
super(PDGD, self).__init__(*args, **kargs)
self.learning_rate = learning_rate
self.learning_rate_decay = learning_rate_decay
self.model = LinearModel(n_features=self.n_features,
learning_rate=learning_rate,
learning_rate_decay=learning_rate_decay,
n_candidates=1)
def __init__(self, alpha, _lambda, refine, rank, update, learning_rate, learning_rate_decay, ind, *args, **kargs):
super(PairRank, self).__init__(*args, **kargs)
self.alpha = alpha
self._lambda = _lambda
self.refine = refine
self.rank = rank
self.update = update
self.learning_rate = learning_rate
self.learning_rate_decay = learning_rate_decay
self.ind = ind
self.A = self._lambda * np.identity(self.n_features)
self.InvA = np.linalg.pinv(self.A)
self.model = LinearModel(
n_features=self.n_features, learning_rate=learning_rate, learning_rate_decay=1, n_candidates=1,
)
self.history = {}
self.n_pairs = []
self.pair_index = []
self.log = {}
self.get_name()
def __init__(self,
k_initial,
k_increase,
n_candidates,
prev_qeury_len=None,
docspace=[False, 0],
*args,
**kargs):
super(P_MGD_DSP, self).__init__(*args, **kargs)
self.n_candidates = n_candidates
self.model = LinearModel(n_features=self.n_features,
learning_rate=self.learning_rate,
n_candidates=self.n_candidates)
self.k_initial = k_initial
self.k_increase = k_increase
self.prev_qeury_len = prev_qeury_len # queue size of features from previous queries
if prev_qeury_len:
self.prev_feat_list = []
# for document space length experiment
# docspace=[True,3] means use superset of document space with three additional documents to perfect DS user examined.
self.docspace = docspace
def __init__(self, n_candidates, *args, **kargs):
super(TD_MGD, self).__init__(*args, **kargs)
self.model = LinearModel(n_features = self.n_features,
learning_rate = self.learning_rate,
n_candidates = n_candidates,
learning_rate_decay = self.model.learning_rate_decay)