issues
search
ai-se
/
LDAClassification
Tuning LDA
https://github.com/amritbhanu/LDADE-package
0
stars
0
forks
source link
Weekly - 2016-11-08
#11
Closed
amritbhanu
closed
7 years ago
amritbhanu
commented
8 years ago
DOING:
Working on these todo lists #6
Will work on Naive Bayes
Find some applications where users can argue with LDA conclusions.
Done
For LN: Mutate Tags by x% to see what error it throws for ambiguity. [Results] (
https://github.com/ai-se/e-disc/issues/106
)
Validation on LDA: tf+l2norm better than LDA better than bag of words with (train, grow, and test sets) #10
Running:
For e-disc, if we dont have labels, can we fix k (100) and only tune alpha beta just for better cluster stability (Tune for jaccard score)?
roadblocks:
Nothing yet
admin:
No
DOING:
Done
Running:
roadblocks:
admin: