def __call__(self):
# delegate to cython implementation
u, i, j, N, trials = warp_sample(self.decomposition.U,
self.decomposition.V, self.train.data,
self.train.indices, self.train.indptr,
self.positive_thresh, self.max_trials,
self.selected_items)
return u, i, j, N, trials #* self.sample_item_rate
def sample(self,train,decomposition):
# delegate to cython implementation
return warp_sample(decomposition.U,
decomposition.V,
train.data,
train.indices,
train.indptr,
self.positive_thresh,
self.max_trials)
def __call__(self):
# delegate to cython implementation
return warp_sample(self.decomposition.U,
self.decomposition.V,
self.train.data,
self.train.indices,
self.train.indptr,
self.positive_thresh,
self.max_trials,
self.selected_items)
def compute_updates(self, train, decomposition, updates):
updates.clear()
tot_trials = 0
###THIS LINE
for ix in xrange(self.batch_size):
###THIS LINE
i, yy, N, trials = warp_sample(decomposition.U, decomposition.V,
train, self.max_trials)
tot_trials += trials
L = self.estimate_warp_loss(train, yy, N)
###THIS LINE
updates.set_update(ix,
decomposition.compute_gradient_step(u, i, j, L))
return tot_trials
def sample(self, train, decomposition):
# delegate to cython implementation
return warp_sample(decomposition.U, decomposition.V, train.data,
train.indices, train.indptr, self.positive_thresh,
self.max_trials)
def compute_updates(self, train, decomposition, updates):
updates.clear()
warp_sample(decomposition.U, decomposition.V, train.data,
train.indices, train.indptr, self.positive_thresh,
self.max_trials)