hpandeycodeit commented on issue #432: MADLIB-1351 : Added stopping criteria on perplexity to LDA URL: https://github.com/apache/madlib/pull/432#issuecomment-548876785 @fmcquillan99, In `lda_predict ` although the model table remains the same, it randomly initializes the output table. That is why we are seeing the difference in the perplexity values from what is calculated in `lda_train` vs `get_perplexity()` However, if the same output table(generated by `lda_train`) is passed to the `get_perplexity()` function, the perplexity values match. For eg: ``` DROP TABLE IF EXISTS lda_model_perp, lda_output_data_perp; SELECT madlib.lda_train( 'documents_tf', -- documents table in the form of term frequency 'lda_model_perp', -- model table created by LDA training (not human readable) 'lda_output_data_perp', -- readable output data table 385, -- vocabulary size 5, -- number of topics 10, -- number of iterations 5, -- Dirichlet prior for the per-doc topic multinomial (alpha) 0.01, -- Dirichlet prior for the per-topic word multinomial (beta) 1, -- Evaluate perplexity every n iterations .2 -- Stopping perplexity tolerance ); ``` Generates the following perplexity values with the last perplexity value **179.380131412**: ``` postgres=# select perplexity from lda_model_perp ; perplexity ---------------------------------------------------------------------------------------------------------------------------------------------- {196.940707618,193.245742228,191.155602156,185.314159394,182.901929923,187.283749958,186.944341124,185.508311039,185.72038473,179.380131412} (1 row) ``` Now running the `get_perplexity()` on the above-generated output table `lda_output_data_perp` produces the following perplexity: ``` postgres=# SELECT madlib.lda_get_perplexity( 'lda_model_perp', postgres(# 'lda_output_data_perp' postgres(# ); lda_get_perplexity -------------------- 179.380131412469 ``` which matches the last perplexity value calculated by `lda_train` Thanks!
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
