⬛ Python Individual Conditional Expectation Plot Toolbox

Last update: Dec 30, 2022

Overview

⬛ PyCEbox

Python Individual Conditional Expectation Plot Toolbox

A Python implementation of individual conditional expecation plots inspired by R's ICEbox. Individual conditional expectation plots were introduced in Peeking Inside the Black Box: Visualizing Statistical Learning with Plots of Individual Conditional Expectation (arXiv:1309.6392).

Quickstart

pycebox is available on PyPI and can be installed with pip install pycebox.

The tutorial recreates the first example in the above paper using pycebox.

Development

For easy development and prototyping using IPython notebooks, a Docker environment is included. To run an IPython notebook with access to your development version of pycebox, run PORT=8889 sh ./start_container.sh. A Jupyter notebook server with access to your development version of pycebox should be available at http://localhost:8889/tree.

To run the pycebox's tests in your development container

Access a bash shell on the container with docker exec -it pycebox bash.
Change to the pycebox directory with cd ../pycebox
Run the tests with pytest test/test.py

Documentation

For details of pycebox's API, consult the documentation.

License

This library is distributed under the MIT License.

Comments

Typo in ice_plot() regarding _get_quantiles()

There is a typo in the ice_plot() function when calling the _get_quantiles() function. In lines 124 and 137, the ice_plot() calls __get_quantiles() (which is undefined) instead of _get_quantiles(), which results in an error if trying to use quantiles or center the ICE curves.

opened by savvastj 6
Using predicted probabilities for binary classification

Is there any way to give some form of predict_proba function to the ice() function in order to see the probability as opposed to the prediction?

Thanks! Nema

opened by nemasobhani 1
Plot mistake

There is a problem in the visualization part. When I am trying to plot the graph in the example, I see the following mistake:

TypeError Traceback (most recent call last) in 12 ice_plot(ice_df, frac_to_plot=0.1, 13 color_by='x3', cmap=PuOr, ---> 14 ax=ice_ax); 15 16 ice_ax.set_xlabel('$X_2$');

C:\ProgramData\Anaconda3\lib\site-packages\pycebox\ice.py in ice_plot(ice_data, frac_to_plot, plot_points, point_kwargs, x_quantile, plot_pdp, centered, centered_quantile, color_by, cmap, ax, pdp_kwargs, **kwargs) 128 if frac_to_plot < 1.: 129 n_cols = ice_data.shape[1] --> 130 icols = np.random.choice(n_cols, size=frac_to_plot * n_cols, replace=False) 131 plot_ice_data = ice_data.iloc[:, icols] 132 else:

mtrand.pyx in mtrand.RandomState.choice()

TypeError: 'float' object cannot be interpreted as an integer

opened by karakol15 4
"frac_to_plot" parameter in ice_plot

Hey Austin,

This package rocks, thanks for publishing it!

I have a question and a potential small bug in the ice_plot method, specifically on the "frac_to_plot" parameter.

It is my understanding that you simply take the fraction and multiply by the number of columns, and then pass this to the "size" parameter of np.random.choice(). I think we should make sure that the number being passed is an integer, not a float. Otherwise np.random.choice() will not accept a float as a parameter for "size".

Current: icols = np.random.choice(n_cols, size=frac_to_plot * n_cols, replace=False)

Fix: icols = np.random.choice(n_cols, size=int(frac_to_plot * n_cols), replace=False)

Best, Andrew

opened by andrew-cho 1
Extended use to classification models, fixed typecast bug
Extended use to classification models by allowing predict_proba to be passed to the ice_plot function.

Fixed 'type error when size is non-int' error for np.random.choice function
opened by sanjifr3 0
Averaging ICE plots across multiple runs/folds of a model

Hi Austin,

I was wondering if it is possible to average across multiple runs/folds of the same model.

I am trying at the moment, but the resulting ICE plots do not make sense. The per run plots make sense but when I average them across both runs and folds the data gets screwed.

Cheers,

Dan

opened by danieltudosiu 0

Releases(0.0.1)

0.0.1(Jan 8, 2017)

Source code(tar.gz)
Source code(zip)

Owner

Austin Rochford

Chief Data Scientist @ Kibo Commerce, recovering mathematician, enthusiastic Bayesian

GitHub Repository

⬛ Python Individual Conditional Expectation Plot Toolbox

Related tags

Overview

⬛ PyCEbox

Quickstart

Development

Documentation

License

Comments

Typo in ice_plot() regarding _get_quantiles()

Using predicted probabilities for binary classification

Plot mistake

"frac_to_plot" parameter in ice_plot

Extended use to classification models, fixed typecast bug

Averaging ICE plots across multiple runs/folds of a model

Releases(0.0.1)

0.0.1(Jan 8, 2017)

Owner

Austin Rochford

GNNLens2 is an interactive visualization tool for graph neural networks (GNN).

python partial dependence plot toolbox

A python library for decision tree visualization and model interpretation.

Convolutional neural network visualization techniques implemented in PyTorch.

⬛ Python Individual Conditional Expectation Plot Toolbox

Making decision trees competitive with neural networks on CIFAR10, CIFAR100, TinyImagenet200, Imagenet

Model analysis tools for TensorFlow

A game theoretic approach to explain the output of any machine learning model.

Lime: Explaining the predictions of any machine learning classifier

Contrastive Explanation (Foil Trees), developed at TNO/Utrecht University

Quickly and easily create / train a custom DeepDream model

Portal is the fastest way to load and visualize your deep neural networks on images and videos 🔮

tensorboard for pytorch (and chainer, mxnet, numpy, ...)

An Empirical Review of Optimization Techniques for Quantum Variational Circuits

A collection of infrastructure and tools for research in neural network interpretability.

pytorch implementation of "Distilling a Neural Network Into a Soft Decision Tree"

A Practical Debugging Tool for Training Deep Neural Networks

🎆 A visualization of the CapsNet layers to better understand how it works

Pytorch implementation of convolutional neural network visualization techniques

Code for "High-Precision Model-Agnostic Explanations" paper