Infinite loop in specific conditions

https://github.com/interpretml/DiCE/blob/a772c8d4fcd88d1cab7f2e02b0bcc045dc0e2eab/dice_ml/explainer_interfaces/explainer_base.py#L73

The post-hoc sparsity enhancement can enter in an infinite loop in certain conditions (I tested with all numerical data, with data ranging, mostly, from -1 to 1). If you need, I can attach here the data. The cause is, for small numbers, sometimes the rate of change is very small or even zero, then, it stucks in this step.

I think one possible solution is to check the difference between the new diff and old difference old_diff, something like:

        if (old_diff == diff) or ((old_diff-diff) < 1e-5):
            n_tolerance += 1
        else:
            n_tolerance = 0

and adding another stop condition on the while loop: n_tolerance <= max_tolerance

[raam93]

Yes, that’s a good catch! Indeed, it gets stuck in an infinite loop for values much lower than zero. However, the default sparsity correction method that you are using performs a linear search. If the option posthoc_sparsity_algorithm in generate_counterfactuals() is changed to “binary”, which performs a binary search instead, the results are much faster. We have observed the binary option work well in practice with few real-world datasets. Can you try it and let me know how it goes?

That said, for the linear method, I really like your suggestion! Can you please raise a PR on the same?

[rmazzine]

Sorry for the late response!

Here’s the updated notebook reproducing the problem: https://colab.research.google.com/drive/1vIJ79cV1WmNAl2usZJIBNIxdcFQn3H_h?usp=sharing

I think the solution can be simpler. From my tests, the linear search should not take more than 10k steps. Therefore, just adding a counter and establishing an upper limit is sufficient.

You can see this change in this commit: https://github.com/rmazzine/DiCE/commit/22f2c12d822a7bcb26a5b1f169e601b217c022cb

And in this notebook below you can see it fixes the problem: https://colab.research.google.com/drive/1W6nlXucrure8hNwQ2bByjxdlxmtF-i3r?usp=sharing

This PR (https://github.com/interpretml/DiCE/pull/293) does that. Therefore, I implement an additional condition for the number of steps that can be changed by the user by modifying the parameter limit_steps_ls.