Skip to content

Latest commit

 

History

History
16 lines (9 loc) · 840 Bytes

README.md

File metadata and controls

16 lines (9 loc) · 840 Bytes
Raw

task_word_explainability

Code for: S Agarwal, YR Semenov, W Lotter. Representing visual classification as a linear combination of words. Proceedings of the 3rd Machine Learning for Health symposium, PMLR (2023).

methods_fig

Set-up/Requirements

  1. Install CLIP and its dependencies by following instructions at: https://github.com/openai/CLIP
  2. Install packages in requirements.txt in order to run all scripts. The code should generally be version agnostic for these packages.

Usage

fit_words.py: Illustrates fitting a linear classifier based on CLIP image embeddings and subquently estimating the classifier based on a linear combination of word embeddings.

plotting.py: Plots the regression word weights.