jdb110 / plda Goto Github PK

What steps will reproduce the problem?
1. checkout svn source code
2. read INSTALL file
3. make OR make all, error: "make: mpicxx：命令未找到"

INSTALL file,
"
Single Processor Version:
  * tar xvfz plda-3.0.tar.gz
  * cd plda
  * make all
"
does not need install  mpich2.
in Makefile:
"all: lda infer mpi_lda"
mpi_lda under target 'all'.

I have not check tar package.

What steps will reproduce the problem?
1. do as the note "quick start" tell me
2. after I copy the sample code and finish executing, no lda-model.txt.
3. I think it is not been created.

What is the expected output? What do you see instead?
I want to see the TXT file, nothing happens.
What version of the product are you using? On what operating system?
3.1,fedro 11

Please provide any additional information below.
there is information of iterations printed on the screen.

I think in document.cc, function

bool LDADocument::WordOccurrenceIterator::Done()

returns true at the first occurrence of the last word, but that word could have 
multiple occurrences. For example, if the last word "zebra" has 3 occurrences, 
the iterator will not get to the last 2.

What steps will reproduce the problem?
1. Use a word count file with an empty line somewhere (i.e. an document that is 
empty for some reason)
2. Call infer on it.

What is the expected output? What do you see instead?
I expect the inferred topics to be in the same line as the word counts.
If there is an empty line, I expect the inference to be all zeros for this 
line, or at least an empty line as well. 
Instead, my output has one line less, which is unfortunate since document 
identifiers are not included in the format used by plda.

What version of the product are you using? On what operating system?
I tried with 2.0 and 3.1 and it occured both times (Ubuntu 11.04).

What steps will reproduce the problem?
1. I read plda-3.1.tar.gz source code ,and fix one memory error in train big 
data project(more than 5000 topics && 260000 terms). How to submit patch?

What is the expected output? What do you see instead?


What version of the product are you using? On what operating system?


Please provide any additional information below.

It looks like the distribution computation in 
LDASampler::GenerateTopicDistributionForWord has:

    distribution->push_back(
        (topic_word_factor + beta_) *
        (document_topic_factor + alpha_) /
        (global_topic_factor + num_topics * beta_));

But based on Algorithm 3.1 in 
http://www.datalab.uci.edu/papers/fastLDA_2008.pdf, shouldn't the 
denominator in the value above be (global_topic_factor + num_words * 
beta_)?

What steps will reproduce the problem?
1. i use a model file that is really big 4000 concepts, 20 000 words 40 000 
articles
2. i try to run infer
3. failes

What is the expected output? What do you see instead?
i get this error message

terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
Aborted

What version of the product are you using? On what operating system?
3.0

Please provide any additional information below.

The wiki page assumes the MPI daemon mpd. However, since MPICH 3, there is no 
longer such daemon.

I provided this installation script for download/build/deployment of MPICH 3.1 
and pLDA 3.1:

https://gist.github.com/wangkuiyi/9922543

Thanks for the diffs

In addition to your changes, I had to add the following to mpi_lda.cc:
#include <algorithm>

Hey I have taken your library and created the infrastructure for a nice little 
web service here http://mnesia.wikispaces.com/Organixe .I thought you might be 
interested...

-Sid