alibi
alibi copied to clipboard
Algorithms for explaining machine learning models
Hi, I had this weird issue that whenever I constructed an explainer, my scripts would stop printing errors. Turns out that `sys.stderr` disappears when calling `cf = CounterFactual(cnn, shape=(1, 28,...
Our `IntegratedGradients` method is for TensorFlow models only, for feature parity it would be desirable to extend it to PyTorch models as well. There are a few options and some...
PDP and accompanying ICE are popular "simple" explanation methods. We have so far not implemented them in `alibi` and providing the more faitfhul `ALE` as an alternative. However, PDP/ICE are...
With `ray` we can optionally run `KernelShap` explanations in parallel as documented [here](https://docs.seldon.io/projects/alibi/en/latest/methods/KernelSHAP.html#Running-batches-of-explanations-in-parallel), but it would be nice to have an example and adding some benchmarks (we already have some...
We may need to clean-up test organization. Some general guidlines: - maintanable fixture files, e.g. currently `alibi.explainers.conftest` has lots of re-usable fixtures, but they can't be used outside of the...
Currently all then`fetch_` dataset functions only store the data in memory, it would be useful if a local copy is saved and used on subsequent calls.
I am using the CounterFactualProto to explain the prediction of a keras LSTM model. I have a input shape of (batch size, timestep, attributes) = (batch size, 86, 32). As...
Hello, I'm new to alibi and am loving it so far! I've been going through the Counterfactual Proto examples and wanted to ask how I could specify which input features...
I'm trying to run `plot_ale` like this: ``` proba_ale = ale_analysis(model, train_dataframe, features_list) plot_ale(proba_ale, n_cols=2, fig_kw={'figwidth': 12, 'figheight': 15}, sharey=None); ``` Here, `model` is my trained model, `train_dataframe` is the...
Now that GPU support is integrated in XGBoost and upstream SHAP (waiting for a new release), we should test that the Alibi integration works as expected. Link to paper: https://arxiv.org/abs/2010.13972