Scale of bar graph does not change

[data-compute-geek]

I am trying to use pyLDAvis to visualize the result of a topic modeling project. I notice that the scale of the axis for the frequency bar does not change when the length of the bars change as I slide the lambda dial. The visualization can be accessed here https://seanlog.com/tmp/lda_visualization_tutorial.html#topic=4&lambda=0&term=

For example, for topic 4, the frequency of christian is >2,000 when lambda is set at 0, and changes to ~ 1,500 when lambda is changed to 0.2.


My understanding is that changing the lambda only changes the order of the word, but not the overall or topic-specific frequencies. Shouldn't the scale of the axis for the bars change as lambda changes so as to keep the frequencies constant?

I used pyLDAvis.gensim_models.prepare to generate the visualization, and the code used to generate the visualization is available at https://github.com/data-compute-geek/pyLDAvis/blob/main/topic-modeling_tutorial.py

[chschroeder]

Same here. I am using pyLDAvis within a notebook:

import pyLDAvis
import pyLDAvis.sklearn

pyLDAvis.enable_notebook()

pyLDAvis.sklearn.prepare(topic_model, x_train, vectorizer)

where topic_model is an instance of sklearn.decomposition.LatentDirichletAllocation and vectorizer is a CountVectorizer.

[ldong87]

This happens to me too. The frequency ticks are now static somehow.