This repository provides notebooks with step-by-step tutorials for implementing concept-based methods for explaining a deep neural networks' internal representations in human-understandable terms.
- Testing with Concept Activation Vectors (TCAV) by Kim et al. (2018).
- Automatic Concept-based Explanation (ACE) by Ghorbani et al. (2019).
- Completeness-aware concept-based explanations by Yeh et al. (2019)
- Causal Concept Effect (CaCE) by Goyal et al. (2020)
- {More to come...}
- Make a copy of this repository in your local machine.
git clone https://github.com/martinagvilas/concept-based-explanation.git
cd concept-based-explanation
- Install the required software to run the notebooks. We recommend that you use miniconda for the installation:
conda create --name cbe python=3.8
conda activate cbe
pip install -r requirements.txt