Lda difference #1334
Lda difference #1334
Conversation
gensim/models/ldamodel.py
Outdated
| z[topic1][topic2] = distance_func(d1[topic1], d2[topic2]) | ||
|
|
||
| if np.abs(np.max(z)) > 1e-8: | ||
| z /= np.max(z) |
There was a problem hiding this comment.
Please comment on when normalisation is important. please make a flag
|
Need to add a notebook with simple tutorial with usage & it's PR will be complete |
| @@ -532,6 +532,10 @@ def jaccard(vec1, vec2): | |||
| return 1 - float(len(intersection)) / float(len(union)) | |||
|
|
|||
|
|
|||
| def jaccard_set(set1, set2): | |||
| return 1. - float(len(set1 & set2)) / float(len(set1 | set2)) | |||
There was a problem hiding this comment.
Will throw an exception if both inputs empty -- is that desired?
Missing docstring.
| >>> print(annotation) # get array with positive/negative words for each topic pair from `m1` and `m2` | ||
| """ | ||
|
|
||
| distances = {"kulback_leibler": kullback_leibler, |
| if np.abs(np.max(z)) > 1e-8: | ||
| z /= np.max(z) | ||
|
|
||
| annotation = [[None for _ in range(t1_size)] for _ in range(t2_size)] |
There was a problem hiding this comment.
You can create lists using *: [None] * t1_size.
Although I don't see the point of this initialization. Why not just start empty and append, in the loop below? What's with the Nones?
There was a problem hiding this comment.
Why not just start empty and append, in the loop below?
Initialization allows writing more readable code (only assignment to the cell in a cycle).
There was a problem hiding this comment.
I see. If that's your worry, isn't creating the 2D matrix as a numpy matrix (2D array) simpler/more readable?
There was a problem hiding this comment.
Numpy matrix with complex object type of element [str, str] is not the best choice
|
|
||
| class TestLdaDiff(unittest.TestCase): | ||
| def setUp(self): | ||
| texts = [['human', 'interface', 'computer'], |
Only first case from PR 1243 integrated into LdaModel as 'diff' method with basic tests