def runBatchEP2(workQueue, lockingPost):
count = 0
for (W, K, docs, alpha, maxiters, thr,
useNewton) in iter(workQueue.get, "none"):
#read current parameter value (and copy)
lam = copy.deepcopy(lockingPost.value())
#calculate correction
ep2 = ep2_lda.EP2_LDA(W, K, docs, alpha, lam, useNewton)
lam = ep2.train(maxiters, thr)
ss = (lam - ep2._eta)
#apply correction
lockingPost.increment(ss)
workQueue.task_done()
count += 1
print("\tdone " + str(count))
def runBatchEP2(W, K, docs, alpha, lam, maxiters, threshold, useNewton):
ep2 = ep2_lda.EP2_LDA(W, K, docs, alpha, lam, useNewton)
lam = ep2.train(maxiters, threshold)
return lam
def update_lambda(self, docs):
ep2 = ep2_lda.EP2_LDA(self._W, self._K, docs, self._alpha, self._lambda, self._useNewton)
self._lambda = ep2.train(self._maxiters, self._threshold)
return (self._alpha, self._lambda)
