Janis Klaise

Scikit-learn model support

2

Is supporting scikit-learn models (even if they are usually not Deep Learning models) on the roadmap? If so, what are the minimum requirements to be implemented for support?

Python deployed example fix

5

Updated the Python deployed model example to respect the API changes for both Skater and Lime since the example was written. Also replaced the binary trained model file with a...

Error initializing Interpretation

2

Currently the example code doesn't work without passing a valid data format at initialization: ``` from skater.core.explanations import Interpretation interpreter = Interpretation() ``` ``` DataSetError Traceback (most recent call last)...

Config system as a separate package

8

I feel like the config system together with the function registry is a killer feature of Thinc, so much so that I can see it as a standalone solution to...

Integrated gradients on raw text PoC

1

This is a WIP PR addressing #289. For applying integrated gradients on raw text, I've written a new class `IntegratedGradientsText` which takes an additional object of type `TextTransformer`. The `TextTransformer`...

Counterfactual refactor v1

10

Builds on top of the old `interop-refactor` branch. Goals: - (v0): Test and replace the current `counterfactual.py` module with the new modular implementation on this branch whilst simplifying some of...

Runtime type checking with beartype

27

Partially addresses #488 by using `beartype` to do runtime type checking. Currently annotated part of the public API (specifically, explainers in the `alibi.explainers` subpackage and methods `__init__`, `fit`, `explain`). The...

Pydantic runtime type validation

2

In light of https://github.com/SeldonIO/alibi/pull/511#issuecomment-950861810 let's test `pydantic` approach to runtime type validaton.

Clarify usage of magic numbers in the KL-LUCB bandit algorithm

2

The constants come from PAC lower bounds proved in the paper: http://proceedings.mlr.press/v30/Kaufmann13.html

Support more diverse input types for methods working on tabular data

12

## Current status Currently tabular data in methods such as `AnchorTabular` and `CounterfactualProto` are expected to be in one of a set of restricted formats, e.g. - `np.ndarray` of homogeneous...