Introduction What Kinds of explanation does CIU generate? How does CIU work? A toy example: predicting breast Permutation feature importance Decision Tree Classifier Random Forest Classifier Gradient Boosting Classifier Explaining a single observation Generating Textual Explanations Drawbacks Introduction Explainablity is a very hot topics in the machine learning research community these days. over the past few years, many methods have been introduced to just understand how a machine learning model makes prediction.
In the 4th week of the Tidy Tuesday project, a very interesting and fun dataset was proposed to the data science community. The dataset contains information about thousands of songs on Spotify’s platform and along with their metadata and audio features. You can download the dataset can using the following piece of code.
4th week of the Tidy Tuesday project
spotify_songs <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-01-21/spotify_songs.csv') head(spotify_songs) ## # A tibble: 6 x 23 ## track_id track_name track_artist track_popularity track_album_id ## <chr> <chr> <chr> <dbl> <chr> ## 1 6f807x0~ I Don't C~ Ed Sheeran 66 2oCs0DGTsRO98~ ## 2 0r7CVbZ~ Memories ~ Maroon 5 67 63rPSO264uRjW~ ## 3 1z1Hg7V~ All the T~ Zara Larsson 70 1HoSmj2eLcsrR~ ## 4 75Fpbth~ Call You ~ The Chainsm~ 60 1nqYsOef1yKKu~ ## 5 1e8PAfc~ Someone Y~ Lewis Capal~ 69 7m7vv9wlQ4i0L~ ## 6 7fvUMiy~ Beautiful~ Ed Sheeran 67 2yiy9cd2QktrN~ ## # .
In my last post, I shared my notes from two talks at Explainable Data Science summer school in Luxembourg. Although every talk in the summer school was interesting and taught me new things but I particularly liked the “Explanations in Philosophy and Psychology” talk by Christos Bechlivanidis. I learned a lot of new things from this this talk specially because what I had focused by them was mainly about the more algorithmic aspect of explainability.
Last September, I had the opportunity to participate in the EXPLAINABLE DATA SCIENCE summer school in Kirchberg, Luxembourg. the summer school was organized by the European Association for Data Science (EuADS) and was held during 10-13 September.
What I specifically liked about this summer school ( of course besides enjoying the the beautiful city of Luxembourg ) was the fact that it covered a vast variety of topics in the explainable machine learning (AI) literature, ranging from visualization, XAI techniques, causality to psychological aspects of explainability.