Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) [ICML 2018]

Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, Rory Sayres

ICML Paper: https://arxiv.org/abs/1711.11279

What is TCAV?

Testing with Concept Activation Vectors (TCAV) is a new interpretability method to understand what signals your neural networks models uses for prediction.

What's special about TCAV compared to other methods?

Typical interpretability methods show importance weights in each input feature (e.g, pixel). TCAV instead shows importance of high level concepts (e.g., color, gender, race) for a prediction class - this is how humans communicate!

Typical interpretability methods require you to have one particular image that you are interested in understanding. TCAV gives an explanation that is generally true for a class of interest, beyond one image (global explanation).

For example, for a given class, we can show how much race or gender was important for classifications in InceptionV3. Even though neither race nor gender labels were part of the training input!

Cool, where do these concepts come from?

TCAV learns concepts from examples. For instance, TCAV needs a couple of examples of female, and something not female to learn a "gender" concept. We have tested a variety of concepts: color, gender, race, textures and many others.

Why use high level concepts instead of input features?

Humans think and communicate using concepts, and not using numbers (e.g., weights to each feature). When there are lots of numbers to combine and reason about (many features), it becomes harder and harder for humans to make sense of the information they are accounting for. TCAV instead delivers explanations in the way humans communicate to each other.

The consumer of the explanation may not know machine learning too well. Can they understand the explanation?

Yes. TCAV is designed to make sense to everyone - as long as they can understand the high level concept!

Sounds good. Do I need to change my network to use TCAV?

No. You don't need to change or retrain your network to use TCAV.

Installation

Tensorflow must be installed to use TCAV. But it isn't included in the TCAV pip package install_requires as a user may wish to use it with either the tensorflow or tensorflow-gpu package. So please pip install tensorflow or tensorflow-gpu as well as the tcav package.

pip install tcav

Requirements

See requirements.txt for a list of python dependencies used in testing TCAV. These will all be installed during pip installation of tcav with the exception of tensorflow, as mentioned above.

How to use TCAV

See Run TCAV.ipynb for step by step guide, after pip installing the tcav package.

mytcav = tcav.TCAV(sess,
                   target,
                   concepts,
                   bottlenecks,
                   act_gen,
                   alphas,
                   cav_dir=cav_dir,
                   num_random_exp=2)

results = mytcav.run()

TCAV for discrete models

We provide a simple example of how to run TCAV on models trained on discrete, non-image data. Please see

cd tcav/tcav_examples/discrete/

You can also find a Jupyter notebook for a model trained on KDD99 in here:

tcav/tcav_examples/discrete/kdd99_discrete_example.ipynb.

Requirements

• tensorflow
• numpy
• Pillow
• matplotlib
• scikit-learn
• scipy

How to run unit tests

python -m tcav.cav_test

python -m tcav.model_test

python -m tcav.tcav_test

python -m tcav.utils_test

How to create a new version of the pip package

1. Ensure the version in setup.py has been updated to a new version.
2. Run python setup.py bdist_wheel --python-tag py3 and python setup.py bdist_wheel --python-tag py2.
3. Run twine upload dist/* to upload the py2 and py3 pip packages to PyPi.