Tomas
Tomas
Based on https://github.com/TeamHG-Memex/eli5/pull/315 and https://github.com/TeamHG-Memex/eli5/pull/329, we use [Grad-CAM](https://arxiv.org/pdf/1610.02391.pdf) to highlight parts of text that contribute to a prediction of a Keras classifier. For example, we can use a call similar...
This PR explains image and text classifiers built in PyTorch using the [Grad-CAM](https://arxiv.org/abs/1610.02391) method, building on #315 and #325. Images example: Using the pretrained `mobilenet_v2` network from `torchvision` and calling...
Hi, I would like to work on ELI5 as a GSOC student again. I would like to merge the Grad-CAM PR's https://github.com/TeamHG-Memex/eli5/pull/325 and https://github.com/TeamHG-Memex/eli5/pull/327, work on extra features like adding...
When explaining scikit-learn estimators we can pass arguments like `vec` or `vectorized` to vectorize raw text into arrays. What if we added the same support for Keras Grad-CAM explanations (https://github.com/TeamHG-Memex/eli5/pull/325)?...
The `explain_prediction_keras` function (https://github.com/TeamHG-Memex/eli5/pull/325) takes a `doc` argument that is the input to a Keras model. Currently `doc` must be a numpy array. It would be nice to support other...
The `explain_prediction` function for keras classifiers (https://github.com/TeamHG-Memex/eli5/pull/315, https://github.com/TeamHG-Memex/eli5/pull/329) accepts a `doc` argument. Currently `doc` must have the shape `(1, dims...)` (batch size equal to 1). It would be good to...
The `explain_prediction` function for keras classifiers (https://github.com/TeamHG-Memex/eli5/pull/315) does not yet support the `target_names` parameter. See https://github.com/TeamHG-Memex/eli5/pull/315#discussion_r293318133 for a description of the feature. We should show names of classes in the...
Currently `format_as_image` (https://github.com/TeamHG-Memex/eli5/pull/315) uses `Pillow` for array resizing with interpolation, and `matplotlib` for applying colormaps to grayscale arrays. It would be good to investigate alternative dependencies for these tasks. `matplotlib`...
Currently `explain_prediction_keras` (https://github.com/TeamHG-Memex/eli5/pull/315) supports the `targets` parameter (that is used to explain a particular class, specified by an interger ID). The argument must be a length 1 list of ints,...
Closes #305. * Take ```stability_selection.StabilitySelection```(https://github.com/scikit-learn-contrib/stability-selection) as an argument to ```transform_feature_names```. * Add stability selection tests. * Mention stability selection in the docs. * Register possibly missing dependencies through a wrapper....