Github user hhbyyh commented on the pull request:
https://github.com/apache/spark/pull/7705#issuecomment-125611535
Thank @feynmanliang for picking this up. I hesitated for a long time about
how these functions should be implemented. I know Gensim is using the similar
way, yet from my understanding, log-perplexity has been used as a measurement
for topic modeling long before the birth of online LDA algorithm. I'm not sure
if we can find a more native way to compute it without invoking the inference
in OnlineLDAOptimizer. For LocalLDAModel, it'll be more self-contained and
impartial (for other LDAOptimizers) if we can implement the perplexity and
prediction just from the two distributions and the input documents parameter.
This is just a layman's opinion and may be overcautious, and sorry for
raising a question without providing solutions. I'll ask around to see the
options.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
