❌Explain any Machine Learning Model with LIME ❌NLP Model Interpretability with LIME

[@bryanparis7779]

Does LIME uses Permutation or Perturbation (tweaking feature values) of the input of interest (x) we would like to explain?

[@shaz-z506]

Hi, This is a good video on Lime and the important one too, could you please make a video on using SHAP, how Shapely values are calculated and how we can interpret the result especially on tabular data.

[@DecisionForest]

Hi Shaz, I’m happy you enjoyed it. Actually explainability with SHAP is the next tutorial I’ll be recording so I’ll post it as soon as it’s done.

[@f1l4nn1m]

How would you use LIME on a BiLSTM Keras-based text classification algorithm if the sequences are vectorized and padded?

[@mainaksen9146]

same question, how we can use LIME on text classification based works?

[@f1l4nn1m]

@@mainaksen9146 There’s a specific video on that, from the same author. I don’t have time to look now, but as soon as I can I’ll share the link with you here.

[@mainaksen1]

@@f1l4nn1m Thank you.

[@tanishasharma3665]

I really liked this video! Thank you Just had a single doubt......How are the weights of the features assigned, as in what is the logic behind that? A sheer link would also suffice!

[@ryuzakace]

Read the paper - "Why Should I trust you". Basically, your classifier might be complex/Black Box but LIME selects neighbors of the prediction, which would be linearly separable. So, they could be explained using simpler models and extracting/interpreting weights on features by linear models is easy.