3 ms·
IMHO one of the most overlooked features is that it's the best eXplainable AI out there IMHO. Spam scores are easily understandable and correctable. You can eve
by riedel 3y ago
IMHO one of the most overlooked features is that it's the best eXplainable AI out there IMHO. Spam scores are easily understandable and correctable. You can even build quite digestible nomograms [1] . Still everyone is using decision trees as XAI example while in my experience they are very unstable in quite indigestible after a depth of 3. I also believe Naïve Bayes is quite undervalued. What many people however ignore is that you also need to estimate a density function for continuous variables, which like with many kernel based methods might the rather interesting and sometimes not so naïve part.
[1] https://orange3.readthedocs.io/projects/orange-visual-programming/en/latest/widgets/visualize/nomogram.html https://orange3.readthedocs.io/projects/orange-visual-progra...
- sweezyjeezy 3y agoThe issue with NB for explainability is that the model scores can be very badly calibrated due to the "naive" assumption. You find especially on longer documents, the NB scores basically clump around 0 and 1 due to multiplying a bunch of dependent scores together as if they were independent. This means you can't really use them to assess how 'confident' the model is on its decision. IMO logistic regression, or better fasttext (rank-limited logistic regression) should be the "default" baseline models you should use for NLP classification. They are trained with cross-entropy loss, so the scores are generally fairly well calibrated to an actual confidence score. Moreover since everything is linear (or at least logit-linear), you can do all the same explainability tricks as with NB (a little more effort for fasttext, but it is doable).