🚀 Beyond Deep Learning

[pat-alt]

This is an issue reserved for the TU Delft Student Software Project '23

Currently, CounterfactualExplanations.jl is adapted to Flux.jl: Julia's most popular ecosystem for Deep Learning. The goal for the remainder of the project is to explore possibilities to make the package compatible with a broader suite of predictive models.

In particular, it would be great to add support for 1) deep learning models trained in Python or R and 2) other common machine learning models trained in Julia:

• Use ONNX.jl to handle Python/R torch models #115
• Interface to <: Supervised MLJ models #69

While #115 is an expected outcome, #69 is expected to be quite challenging. In case you hit a major road block here, we may jointly decide to wrap this up early and instead tackle another interesting and challenging problem:

• [Generator] MINT #105

Optionally, you may also want to look at the following issues, should they still be open:

• Dealing with count data #83 (nice-to-have)
• Improve visualisation for multivariate case TaijaPlotting.jl#19
• Improve macros #128