Comments (7)
I don't see a big difference in these examples. For tokens it made it worse. For chars it probably made it more correct. I think it is a good option to have, +1 to add it and enable for char ngrams.
from eli5.
For char_wb and for token ngrams it is not clear what is better.
from eli5.
I agree that it's worse for tokens, but even for char ngrams, the main difference seems to be that bias (the largest feature) is less intense compared to shorter char n-grams, so they look brighter (about the same brightness as bias).
I'd rather have it off by default in all cases, but I'm not entirely sure.
One more questions: I just realised that currently weights that are displayed on mouse over in title are different depending on the preserve_density option value, that is clearly wrong, we should have original weights in title, right?
from eli5.
Let's take 'software' token in the following text as an example:
hi there, i am here looking for some help. my friend is a interior decor designer. he is from thailand. he is trying to find some graphics software on pc. any suggestion on which software to buy,where to buy and how much it costs ? he likes the most sophisticated software(the more features it has,the better)
In char-based example it is light-green. The total weight it gives to the result is quite large though: about 0.1 (a bit less for some of them) for each of char 3- and 4-grams (' sof', 'sof', 'soft', 'oft', 'oftw', 'ftw', 'ftwa', 'twa', 'twar', 'war', 'ware', 'are'); it sums up to a value larger than BIAS. But bias is much brighter here:
It has L value which is 2x brighter than the brightest part of 'software'. In token-based example 'software' is bright-green because it gets a single large weight.
from eli5.
Indeed, you are right, I see the point now!
from eli5.
One more questions: I just realised that currently weights that are displayed on mouse over in title are different depending on the preserve_density option value, that is clearly wrong, we should have original weights in title, right?
After thinking about it more, I think we should display the same weight in title as we use for highlighting, especially in the light of your last example.
from eli5.
Fixed in #32
from eli5.
Related Issues (20)
- Explaining multi-input model HOT 1
- import show_prediction issue HOT 3
- eli5 explains the importance of features by category
- pandas.core.indexes.base.Index not supported for feature_names params in show_weights
- How weights are calculated during local explainations?
- More detailed documentation of eli5.explain_prediction?
- PermutationImportance uses CV splitter indexes incorrectly
- Error: estimator XGBClassifier is not supported
- Making the world a cleaner place
- TypeError: check_cv() takes from 0 to 2 positional arguments but 3 were given HOT 2
- explain_prediction_lightgbm doesnt work for microsoft lightgbm HOT 1
- Pin version of jinja2 (and markupsafe) in requirements
- AttributeError when importing eli5 0.11.0
- Looping with explain_prediction_df crashes
- eli5.show_prediction doesn't work for my custom trained model
- Understanding show_weights html
- Pairing RFECV and RandomForestRegressor with non-cv Permutation Importance returns NotFittedError
- Error importing eli5.lime due to deprecated 'itemfreq' from 'scipy.stats'
- Error : Estimator is not supported
- Does not support scikit-learn=1.3.0 HOT 7
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from eli5.