Speed and power: fine-tuning distilled transformers models

Thu 25 February 2021

Leaner and faster Transformers with distillation. Fine-tuning BERT-tiny in a couple of minutes with PyTorch.


Multi-label classification with BERT

Fri 29 January 2021

Fine-tuning DistillBERT to tag toxic comments on Wikipedia, using TensorFlow and Hugging Face's transformers library.


Understanding the SHAP interpretation method: Kernel SHAP

Sat 29 February 2020

Exploring the mechanics of the SHAP feature attribution method with toy examples. Presentation of Kernel SHAP.


Understanding the SHAP interpretation method: Shapley values

Fri 27 December 2019

Exploring the mechanics of the SHAP feature attribution method with toy examples. Presentation of Shapley values.


Few-shot learning in NLP: many-classes classification from few examples

Sun 19 August 2018

Learning multi-class classification with only a handful of of examples per class may seem impossible at first, let's see how we can tackle this problem with textual data.


LIME of words: interpreting Recurrent Neural Networks predictions

Tue 12 September 2017

Playing with the LIME machine learning model interpretion part 2: deep learning.


LIME of words: how to interpret your machine learning model predictions

Mon 07 August 2017

Playing with the LIME machine learning model interpretation tool.