Fix PR #963 : Passing all the params through the apply call in lda.get_document_topics #978

Partho · 2016-10-26T01:01:27Z

Hi, I have added the test case to check get_document_topics() working on corpus or not. However, the build will fail because of the new test case I have added. For some words in corpus, word_topics and word_phis are not returned .

If the corpus is small, I am getting the correct value. However, if I am taking the corpus as defined in test_ldamodel.py, then it throws the error . Let me show this via a toy example called app.py

from gensim.corpora import mmcorpus, Dictionary
from gensim import models


texts = [['human', 'interface', 'computer'],
['survey', 'user', 'computer', 'system', 'response', 'time'],
['eps', 'user', 'interface', 'system']]
dictionary = Dictionary(texts)
corpus = [dictionary.doc2bow(text) for text in texts]

lda = models.LdaModel(corpus)

# ab = lda.get_document_topics(corpus)
# for t in ab:
#     print('new doc')
#     print(t)
#
# print len(ab)
# print len(corpus)

doc_topics, word_topics, word_phis = lda.get_document_topics(corpus, per_word_topics=True, minimum_probability=0.02)

for t in doc_topics:
    print('new doc_topic')
    print(t)

for t in word_topics:
    print('new word_topic')
    print(t)

for t in word_phis:
    print('new word_phis')
    print(t)

For the above code, I am getting complete output as follows:

new doc_topic
[(96, 0.75249999999999984)]
new doc_topic
[(0, [96]), (1, [96]), (2, [96])]
new doc_topic
[(0, [(96, 0.99999999999999989)]), (1, [(96, 1.0)]), (2, [(96, 1.0)])]
new word_topic
[(2, 0.85857142857143109)]
new word_topic
[(1, [2]), (3, [2]), (4, [2]), (5, [2]), (6, [2]), (7, [2])]
new word_topic
[(1, [(2, 1.0)]), (3, [(2, 1.0)]), (4, [(2, 1.0)]), (5, [(2, 1.0)]), (6, [(2, 1.0)]), (7, [(2, 1.0)])]
new word_phis
[(29, 0.80200000000000238)]
new word_phis
[(0, [29]), (6, [29]), (7, [29]), (8, [29])]
new word_phis
[(0, [(29, 0.99999999999999989)]), (6, [(29, 1.0)]), (7, [(29, 0.99999999999999989)]), (8, [(29, 0.99999999999999989)])]

However, if I take the corpus as:

texts = [['human', 'interface', 'computer'],
['survey', 'user', 'computer', 'system', 'response', 'time'],
['eps', 'user', 'interface', 'system'],
['system', 'human', 'system', 'eps'],
['user', 'response', 'time'],
['trees'],
['graph', 'trees'],
['graph', 'minors', 'trees'],
['graph', 'minors', 'survey']]

then the following error will be thrown:

/Users/partho/anaconda/bin/python /Users/partho/PycharmProjects/untitled/app.py
Traceback (most recent call last):
  File "/Users/partho/PycharmProjects/untitled/app.py", line 39, in <module>
    doc_topics, word_topics, word_phis = lda.get_document_topics(corpus, per_word_topics=True, minimum_probability=0.02)
ValueError: too many values to unpack

Please let me know how to modify the code for word_topics and word_phis in the returned tuple so that it won't throw the ValueError.

This reverts commit 49dce4d.

tmylk · 2016-10-26T07:36:26Z

is it just that they are filtered?

Feel free to change the minimum_phi_value passed

Partho · 2016-10-26T10:33:32Z

Hi @tmylk ,

For the code below :


from gensim.corpora import mmcorpus, Dictionary
from gensim import models


texts = [['human', 'interface', 'computer'],
['survey', 'user', 'computer', 'system', 'response', 'time'],
['eps', 'user', 'interface', 'system'],
['system', 'human', 'system','eps'],
['user', 'response', 'time'],
['trees'],
['graph', 'trees'],
['graph', 'minors', 'trees'],
['graph', 'minors', 'survey']]
dictionary = Dictionary(texts)
corpus = [dictionary.doc2bow(text) for text in texts]

lda = models.LdaModel(corpus)

topics = lda.get_document_topics(corpus, per_word_topics=True)

# for index, topic in enumerate(doc_topics):
#     print len(corpus[index])
#     print doc_topics.corpus[index]


for t in topics:
    print('new doc')
    print 'Document topics:', t[0]
    print 'Word topics:', t[1]
    print 'Word phis:', t[2]
    print('--------------')

Should the output be like this ??

/Users/partho/anaconda/bin/python /Users/partho/PycharmProjects/untitled/app.py
new doc
Document topics: [(80, 0.75250000000000072)]
Word topics: [(0, [80]), (1, [80]), (2, [80])]
Word phis: [(0, [(80, 0.99999999999999989)]), (1, [(80, 0.99999999999999989)]), (2, [(80, 1.0)])]
--------------
new doc
Document topics: [(56, 0.85857142857142965)]
Word topics: [(1, [56]), (3, [56]), (4, [56]), (5, [56]), (6, [56]), (7, [56])]
Word phis: [(1, [(56, 1.0)]), (3, [(56, 1.0)]), (4, [(56, 1.0)]), (5, [(56, 1.0)]), (6, [(56, 1.0)]), (7, [(56, 1.0)])]
--------------
new doc
Document topics: [(93, 0.80200000000000016)]
Word topics: [(0, [93]), (6, [93]), (7, [93]), (8, [93])]
Word phis: [(0, [(93, 1.0)]), (6, [(93, 1.0)]), (7, [(93, 1.0)]), (8, [(93, 1.0)])]
--------------
new doc
Document topics: [(80, 0.80200000000000049)]
Word topics: [(2, [80]), (6, [80]), (8, [80])]
Word phis: [(2, [(80, 1.0)]), (6, [(80, 2.0)]), (8, [(80, 0.99999999999999989)])]
--------------
new doc
Document topics: [(1, 0.75250000000000394)]
Word topics: [(3, [1]), (4, [1]), (7, [1])]
Word phis: [(3, [(1, 1.0)]), (4, [(1, 0.99999999999999989)]), (7, [(1, 1.0)])]
--------------
new doc
Document topics: [(70, 0.50499999999999989)]
Word topics: [(9, [70])]
Word phis: [(9, [(70, 1.0)])]
--------------
new doc
Document topics: [(86, 0.67000000000000048)]
Word topics: [(9, [86]), (10, [86])]
Word phis: [(9, [(86, 1.0)]), (10, [(86, 1.0)])]
--------------
new doc
Document topics: [(4, 0.75250000000000372)]
Word topics: [(9, [4]), (10, [4]), (11, [4])]
Word phis: [(9, [(4, 1.0)]), (10, [(4, 1.0)]), (11, [(4, 1.0)])]
--------------
new doc
Document topics: [(27, 0.75250000000000283)]
Word topics: [(5, [27]), (10, [27]), (11, [27])]
Word phis: [(5, [(27, 0.99999999999999989)]), (10, [(27, 1.0)]), (11, [(27, 1.0)])]
--------------

Process finished with exit code 0

tmylk · 2016-10-27T18:48:32Z

So what is happening in topic 3?

Partho · 2016-10-27T20:06:14Z

At this line, we are returning (document_topics, word_topic, word_phi) as a tuple. So, if you call the following function:
doc_topics, word_topic, word_phis = lda.get_document_topics(corpus, per_word_topics=True), you are basically storing each tuple in a particular variable, i.e., variable doc_topics store tuple data (document_topics, word_topic, word_phi) for document 1, and variables word_topic and word_phis store tuple data for document 2 and 3 respectively. Since,
lda.get_document_topics(corpus, per_word_topics=True) returns a list of tuples and the number of these tuples is more than 3 (corpus size is 9), and 3 variables aren't enough to store these data😃

I will show below for others to realize what mistake they are making while making call to function get_document_topics for the entire corpus:

from gensim.corpora import mmcorpus, Dictionary
from gensim import models


texts = [['human', 'interface', 'computer'],
['survey', 'user', 'computer', 'system', 'response', 'time'],
['eps', 'user', 'interface', 'system']]

dictionary = Dictionary(texts)
corpus = [dictionary.doc2bow(text) for text in texts]

lda = models.LdaModel(corpus)

doc_topics, word_topics, word_phis = lda.get_document_topics(corpus, per_word_topics=True)

print (doc_topics)
print (word_topics)
print (word_phis)

will output :

[[(19, 0.75250000000000328)], [(0, [19]), (1, [19]), (2, [19])], [(0, [(19, 1.0)]), (1, [(19, 1.0)]), (2, [(19, 0.99999999999999989)])]]

[[(86, 0.85857142576872991)], [(1, [86]), (3, [86]), (4, [86]), (5, [86]), (6, [86]), (7, [86])], [(1, [(86, 1.0)]), (3, [(86, 1.0)]), (4, [(86, 1.0)]), (5, [(86, 1.0)]), (6, [(86, 1.0)]), (7, [(86, 1.0)])]]

[[(77, 0.80200000000000071)], [(0, [77]), (6, [77]), (7, [77]), (8, [77])], [(0, [(77, 1.0)]), (6, [(77, 1.0)]), (7, [(77, 1.0)]), (8, [(77, 1.0)])]]

As we can see, each tuple has 3 lists of document_topics, word_topic, word_phi. In case the corpus length is >3, the above code will throw valueError.

So, the proper way to access the data for the entire corpus is like this :

all_topics = lda.get_document_topics(corpus, per_word_topics=True)
for doc_topic, word_topic, word_phi in all_topics:
    print('new doc')
    print 'Document topics:', doc_topic
    print 'Word topics:', word_topic
    print 'Word phis:', word_phi
    print('--------------')

which will return:

new doc
Document topics: [(80, 0.75250000000000072)]
Word topics: [(0, [80]), (1, [80]), (2, [80])]
Word phis: [(0, [(80, 0.99999999999999989)]), (1, [(80, 0.99999999999999989)]), (2, [(80, 1.0)])]
--------------
new doc
Document topics: [(56, 0.85857142857142965)]
Word topics: [(1, [56]), (3, [56]), (4, [56]), (5, [56]), (6, [56]), (7, [56])]
Word phis: [(1, [(56, 1.0)]), (3, [(56, 1.0)]), (4, [(56, 1.0)]), (5, [(56, 1.0)]), (6, [(56, 1.0)]), (7, [(56, 1.0)])]
--------------
new doc
Document topics: [(93, 0.80200000000000016)]
Word topics: [(0, [93]), (6, [93]), (7, [93]), (8, [93])]
Word phis: [(0, [(93, 1.0)]), (6, [(93, 1.0)]), (7, [(93, 1.0)]), (8, [(93, 1.0)])]
--------------
new doc
Document topics: [(80, 0.80200000000000049)]
Word topics: [(2, [80]), (6, [80]), (8, [80])]
Word phis: [(2, [(80, 1.0)]), (6, [(80, 2.0)]), (8, [(80, 0.99999999999999989)])]
--------------
new doc
Document topics: [(1, 0.75250000000000394)]
Word topics: [(3, [1]), (4, [1]), (7, [1])]
Word phis: [(3, [(1, 1.0)]), (4, [(1, 0.99999999999999989)]), (7, [(1, 1.0)])]
--------------
                                              .
                                              .
                                              .

And if we want to get the data for doc_topics like
[(80, 0.75250000000000072), (56, 0.85857142857142965), ...] , and word_topics like
[( 0, [80, 93, ...] ), ( 6, [ 80, 93, ... ] ), ...] , then we will have to add the code to restructure the data and pass it in three variables doc_topic, word_topic, word_phi .

Partho · 2016-10-27T23:04:37Z

In order to get the doc_topics, word_topics and word_phis of the overall corpus, I restructured the output data by creating a helper function. I want to confirm whether the output of the aggregated result is desirable or not. If yes, then we can include it in the final code.

The helper function is as follows:

def get_document_topics_for_corpus(topics):
    document_topics, word_topics, word_phis = dict(), dict(), dict()
    doc_topic_list, word_topic_list, word_phi_list = list(), list(), list()

    for doc_index, (doc_topic, word_topic, word_phi) in enumerate(topics):
        #Document_topics aggregation
        #document_topics.setdefault(doc_topic[0][0], []).append(doc_topic[0][1])
        document_topics.setdefault(doc_index, doc_topic[0])

        #Word_topics aggregation
        for key in word_topic:
            word_topics.setdefault(doc_index, []).append(key)

        #Word_phis aggregation
        for key in word_phi:
            word_phis.setdefault(doc_index, []).append(key)

    doc_topic_list.extend(document_topics.items())
    word_topic_list.extend(word_topics.items())
    word_phi_list.extend(word_phis.items())

    return (doc_topic_list, word_topic_list, word_phi_list)

And now, I can call the above function as follows:

topics = lda.get_document_topics(corpus, per_word_topics=True)
doc_topics, word_topics, word_phis = get_document_topics_for_corpus(topics)

print "Document topics: ", doc_topics
print "Word topics: ", word_topics
print "Word phis:", word_phis

The actual result is as follows:

new doc
Document topics: [(96, 0.75249999999999995)]
Word topics: [(0, [96]), (1, [96]), (2, [96])]
Word phis: [(0, [(96, 1.0)]), (1, [(96, 0.99999999999999989)]), (2, [(96, 0.99999999999999989)])]
--------------
new doc
Document topics: [(83, 0.85857142857142887)]
Word topics: [(1, [83]), (3, [83]), (4, [83]), (5, [83]), (6, [83]), (7, [83])]
Word phis: [(1, [(83, 1.0)]), (3, [(83, 1.0)]), (4, [(83, 1.0)]), (5, [(83, 1.0)]), (6, [(83, 1.0)]), (7, [(83, 1.0)])]
--------------
new doc
Document topics: [(28, 0.80199999999858307)]
Word topics: [(0, [28]), (6, [28]), (7, [28]), (8, [28])]
Word phis: [(0, [(28, 1.0)]), (6, [(28, 1.0)]), (7, [(28, 1.0)]), (8, [(28, 1.0)])]
--------------
new doc
Document topics: [(45, 0.80199999999999594)]
Word topics: [(2, [45]), (6, [45]), (8, [45])]
Word phis: [(2, [(45, 1.0)]), (6, [(45, 2.0)]), (8, [(45, 1.0)])]
--------------
new doc
Document topics: [(40, 0.75250000000000217)]
Word topics: [(3, [40]), (4, [40]), (7, [40])]
Word phis: [(3, [(40, 1.0000000000000002)]), (4, [(40, 1.0)]), (7, [(40, 1.0000000000000002)])]
--------------
new doc
Document topics: [(79, 0.50499999999999989)]
Word topics: [(9, [79])]
Word phis: [(9, [(79, 1.0)])]
--------------
new doc
Document topics: [(79, 0.67000000000000082)]
Word topics: [(9, [79]), (10, [79])]
Word phis: [(9, [(79, 1.0)]), (10, [(79, 1.0)])]
--------------
new doc
Document topics: [(79, 0.75249999997567552)]
Word topics: [(9, [79]), (10, [79]), (11, [79])]
Word phis: [(9, [(79, 0.99999999999999989)]), (10, [(79, 1.0)]), (11, [(79, 1.0)])]
--------------
new doc
Document topics: [(75, 0.75250000000000095)]
Word topics: [(5, [75]), (10, [75]), (11, [75])]
Word phis: [(5, [(75, 1.0)]), (10, [(75, 1.0)]), (11, [(75, 1.0)])]
--------------

And the aggregated result are as follows:

Document topics:  [(0, (93, 0.75250000000000017)), (1, (94, 0.85857142857142865)), (2, (64, 0.80200000000000116)), (3, (97, 0.80199999896254648)), (4, (75, 0.75249999999877881)), (5, (85, 0.50499999999999989)), (6, (51, 0.67000000000000226)), (7, (52, 0.75250000000000183)), (8, (49, 0.75250000000000195))]

Word topics:  [(0, [(0, [93]), (1, [93]), (2, [93])]), (1, [(1, [94]), (3, [94]), (4, [94]), (5, [94]), (6, [94]), (7, [94])]), (2, [(0, [64]), (6, [64]), (7, [64]), (8, [64])]), (3, [(2, [97]), (6, [97]), (8, [97])]), (4, [(3, [75]), (4, [75]), (7, [75])]), (5, [(9, [85])]), (6, [(9, [51]), (10, [51])]), (7, [(9, [52]), (10, [52]), (11, [52])]), (8, [(5, [49]), (10, [49]), (11, [49])])]

Word phis: [(0, [(0, [(93, 1.0)]), (1, [(93, 1.0)]), (2, [(93, 1.0)])]), (1, [(1, [(94, 1.0)]), (3, [(94, 1.0)]), (4, [(94, 1.0)]), (5, [(94, 1.0)]), (6, [(94, 1.0)]), (7, [(94, 1.0)])]), (2, [(0, [(64, 1.0)]), (6, [(64, 1.0)]), (7, [(64, 1.0)]), (8, [(64, 1.0)])]), (3, [(2, [(97, 1.0)]), (6, [(97, 2.0)]), (8, [(97, 1.0)])]), (4, [(3, [(75, 0.99999999999999989)]), (4, [(75, 0.99999999999999989)]), (7, [(75, 0.99999999999999989)])]), (5, [(9, [(85, 1.0)])]), (6, [(9, [(51, 1.0)]), (10, [(51, 1.0)])]), (7, [(9, [(52, 1.0)]), (10, [(52, 1.0)]), (11, [(52, 1.0)])]), (8, [(5, [(49, 1.0)]), (10, [(49, 1.0)]), (11, [(49, 1.0)])])]

Is the aggregated result desirable? If yes, then I will work on improving this helper function. Can you also give me some tips in improving this helper function ??

In case the document topics are unique throughout the corpus (can it happen ?), I can structure the data so that it will look like this:

Document topics:  [(33, 0.80199999999605875), (5, 0.75250000000000383), (39, 0.75250000000000239), (12, 0.50499999998756462), (83, 0.75250000000000061), (52, 0.85857142808535136), (23, 0.67000000000000359), (24, 0.80200000000000249), (90, 0.75250000000000028)]

Word topics:  [(0, [5, 33]), (1, [5, 52]), (2, [5, 24]), (3, [52, 83]), (4, [52, 83]), (5, [52, 90]), (6, [52, 33, 24]), (7, [52, 33, 83]), (8, [33, 24]), (9, [12, 23, 39]), (10, [23, 39, 90]), (11, [39, 90])]

Word phis: [(0, [(5, 1.0), (33, 1.0)]), (1, [(5, 0.99999999999999989), (52, 0.99999999999999989)]), (2, [(5, 0.99999999999999989), (24, 1.0)]), (3, [(52, 1.0), (83, 1.0)]), (4, [(52, 0.99999999999999989), (83, 0.99999999999999989)]), (5, [(52, 1.0), (90, 1.0)]), (6, [(52, 0.99999999999999989), (33, 1.0), (24, 2.0)]), (7, [(52, 0.99999999999999989), (33, 1.0), (83, 0.99999999999999989)]), (8, [(33, 1.0), (24, 1.0)]), (9, [(12, 0.99999999999999989), (23, 1.0000000000000002), (39, 1.0)]), (10, [(23, 1.0), (39, 1.0), (90, 1.0)]), (11, [(39, 1.0), (90, 1.0)])]

tmylk · 2016-10-28T15:41:50Z

I don't see a need for aggregated function - please convince me why you think it is heplful.

The current API below looks ok, though it needs to be documented in the docstring and in the lda ipynb tutorial

all_topics = lda.get_document_topics(corpus, per_word_topics=True)
for doc_topic, word_topic, word_phi in all_topics:

Partho · 2016-10-28T18:39:05Z

Yes, I do agree that the aggregated function is not needed actually, but users were confused and calling like this:

doc_topics, word_topics, word_phis  = lda.get_document_topics(corpus, per_word_topics=True)

instead of this:

topics =  lda.get_document_topics(corpus, per_word_topics=True)

It will be a good idea to be documented in docstring and lda tutorial.

Plus, users can store the data for all corpus like this:

topics =  lda.get_document_topics(corpus, per_word_topics=True)
all_topics = [(doc_topics, word_topics, word_phis) for doc_topics, word_topics, word_phis in topics]

And access a particular document details like this:

print all_topics[doc_index]

and output like this:

[[(19, 0.75250000000000328)], [(0, [19]), (1, [19]), (2, [19])], [(0, [(19, 1.0)]), (1, [(19, 1.0)]), (2, [(19, 0.99999999999999989)])]]

tmylk · 2016-10-31T18:46:33Z

Sounds great. Let's update docstring and tutorial.

Partho · 2016-10-31T19:08:42Z

@tmylk I have updated the tutorial in topic_methods.ipynb . However, the code that I have added in tutorial will only run if we merge the code in this PR. So, can you please review the code supposed to be merged, and let me know what other changes to be made?

tmylk · 2016-11-07T16:05:59Z

gensim/models/ldamodel.py

@@ -977,7 +985,9 @@ def __getitem__(self, bow, eps=None):
        Ignore topics with very low probability (below `eps`).

        """
-        return self.get_document_topics(bow, eps)
+        #Is eps equivalent to minimum_probability?
+        eps = self.minimum_probability


what is the reason to add this line? it is used anyway in https://github.com/parthoiiitm/gensim/blob/2ac2f951452e7d968acace2b6d7b11be8b79ad62/gensim/models/ldamodel.py#L912

tmylk · 2016-11-07T16:15:23Z

gensim/models/ldamulticore.py

@@ -80,7 +80,8 @@ class LdaMulticore(LdaModel):
    def __init__(self, corpus=None, num_topics=100, id2word=None, workers=None,
                 chunksize=2000, passes=1, batch=False, alpha='symmetric',
                 eta=None, decay=0.5, offset=1.0, eval_every=10, iterations=50,
-                 gamma_threshold=0.001, random_state=None):
+                 gamma_threshold=0.001, random_state=None, 


where is minimium_probability? Please add it and pass through

tmylk · 2016-11-07T16:19:16Z

gensim/test/test_ldamodel.py

@@ -249,6 +249,26 @@ def testGetDocumentTopics(self):
                self.assertTrue(isinstance(k, int))
                self.assertTrue(isinstance(v, float))

+        #Test case to use the get_document_topic function for the corpus


Please add a new test for minimum_prob and minimum_phi_value. Select values that change the number of topics and docs returned

tmylk · 2016-11-08T16:16:42Z

gensim/test/test_ldamodel.py

+        #Test case to use the get_document_topic function for the corpus
+        all_topics = model.get_document_topics(self.corpus, minimum_probability=0.5, minimum_phi_value=0.3, per_word_topics=True)
+
+        self.assertEqual(model.state.numdocs, len(corpus))


so the filtering params have no effect? It is a better test if they change the outcome

I am seeing the effect in case I choose minimum_probability=0.8 and minimum_phi_value=1.0, then most of the doc topic list and word phi list will be blank. Should I use that as a test case?

…um_phi_value

Partho Mandal added 4 commits October 25, 2016 17:44

Fix PR #963

df53afa

Added entry in CHANGELOG.md

49dce4d

Revert "Added entry in CHANGELOG.md"

0cc6d3a

This reverts commit 49dce4d.

Added new entry in CHANGELOG.md

102393c

tmylk mentioned this pull request Oct 26, 2016

Passing all the params through the apply call in lda.get_document_topics() #963

Closed

Partho Mandal added 2 commits October 27, 2016 03:59

Update test_ldamodel.py

a82fb57

Update CHANGELOG.md

bf94be5

Updated tutorial for get_document_topics

2ac2f95

tmylk reviewed Nov 7, 2016

View reviewed changes

System Administrator and others added 2 commits November 7, 2016 20:37

Added parameters in ldamulticore constructor

121a2e2

Added parameters for ldamulticore

e01c0a5

tmylk reviewed Nov 8, 2016

View reviewed changes

This was referenced Nov 8, 2016

Resolving issue #908 #942

Closed

Add per_word_topics parameter to apply call #923

Closed

Update test_ldamodel.py

387a6ec

tmylk mentioned this pull request Nov 10, 2016

per_word_topics for a corpus #908

Closed

test added to check filtering effect of minimum_probability and minim…

587ca88

…um_phi_value

tmylk merged commit 17a01cf into piskvorky:develop Nov 11, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix PR #963 : Passing all the params through the apply call in lda.get_document_topics #978

Fix PR #963 : Passing all the params through the apply call in lda.get_document_topics #978

Partho commented Oct 26, 2016 •

edited

Loading

tmylk commented Oct 26, 2016

Partho commented Oct 26, 2016 •

edited

Loading

tmylk commented Oct 27, 2016

Partho commented Oct 27, 2016 •

edited

Loading

Partho commented Oct 27, 2016 •

edited

Loading

tmylk commented Oct 28, 2016 •

edited

Loading

Partho commented Oct 28, 2016 •

edited

Loading

tmylk commented Oct 31, 2016

Partho commented Oct 31, 2016 •

edited

Loading

tmylk Nov 7, 2016

tmylk Nov 7, 2016 •

edited

Loading

tmylk Nov 7, 2016

tmylk Nov 8, 2016 •

edited

Loading

Partho Nov 8, 2016

tmylk Nov 9, 2016

Fix PR #963 : Passing all the params through the apply call in lda.get_document_topics #978

Fix PR #963 : Passing all the params through the apply call in lda.get_document_topics #978

Conversation

Partho commented Oct 26, 2016 • edited Loading

tmylk commented Oct 26, 2016

Partho commented Oct 26, 2016 • edited Loading

tmylk commented Oct 27, 2016

Partho commented Oct 27, 2016 • edited Loading

Partho commented Oct 27, 2016 • edited Loading

tmylk commented Oct 28, 2016 • edited Loading

Partho commented Oct 28, 2016 • edited Loading

tmylk commented Oct 31, 2016

Partho commented Oct 31, 2016 • edited Loading

tmylk Nov 7, 2016

Choose a reason for hiding this comment

tmylk Nov 7, 2016 • edited Loading

Choose a reason for hiding this comment

tmylk Nov 7, 2016

Choose a reason for hiding this comment

tmylk Nov 8, 2016 • edited Loading

Choose a reason for hiding this comment

Partho Nov 8, 2016

Choose a reason for hiding this comment

tmylk Nov 9, 2016

Choose a reason for hiding this comment

Partho commented Oct 26, 2016 •

edited

Loading

Partho commented Oct 26, 2016 •

edited

Loading

Partho commented Oct 27, 2016 •

edited

Loading

Partho commented Oct 27, 2016 •

edited

Loading

tmylk commented Oct 28, 2016 •

edited

Loading

Partho commented Oct 28, 2016 •

edited

Loading

Partho commented Oct 31, 2016 •

edited

Loading

tmylk Nov 7, 2016 •

edited

Loading

tmylk Nov 8, 2016 •

edited

Loading