how should I understand the get new topic ?

[guangbaowan]

Hi, David. How can I understand

                        # Get new topic.
                        topic_distribution = (self.topic_word_counts[:, w_index] + beta) * \
                            (self.document_topic_counts[d_index] + alpha) / \
                            (self.topic_counts + beta) # changed by hitalex

.Thank you.

[srilekha1993]

Hi,David As we know perplexity is a measure to find the model performance.Decreasing the perplexity means our model works better. But i am getting trouble below things

I am trying to run the code when stop word list is empty.i got the my perplexites in iteration0 1097.55 iteration1 1097.11 iteration2 1096.54 iteration2 1095.800 with removing puntuation perplexities are iteration0 1390.025 iteration1is1389.333 iteration2is1388.273 iteration3is1386.40980 iteration4is1383.11679

and removing punctuation and and single length words my perplexities iteration0 1517.622 iteration1 1516.79 iteration2 1515.46 iteration3 1513.18 iteration4 1509.46 without using stop word i got at starting 1097 and then decreasing with removing punctuation i am getting 1390 and then decreasing but it should less than 1097 as we are filtering our data and the same thing for removing puntuations and single length word 1517. can you suggest me anything about my problem?