Miguel de Benito Delgado
Miguel de Benito Delgado
`test_influence_linear_model` takes > 300 sec on my machine, while all other inversion methods take < 10sec and `test_inversion_methods` fails on the bigger problem dimension (5,10), presumably because of LiSSA
* [x] #259 * [ ] Reproduce experiments on all 9 datasets
Currently it does one full permutation before the other one. This is how it's described in the paper which first described antithetic permutation sampling. It's the same as with sets:...
If we keep the batching hack, I think it belongs in the samplers. A simple way to do it is to wrap them in a generic `batched` decorator like [this...
Unlike [`mkdocs-jupyter`](https://github.com/danielfrg/mkdocs-jupyter/blob/main/mkdocs_jupyter/plugin.py#L107), [`mknotebooks`](https://github.com/greenape/mknotebooks/blob/master/mknotebooks/plugin.py#L263) converts the notebooks to markdown and lets mkdocs handle the conversion from markdown to HTML. That makes it work with other plugins and extensions. That's why those...
What about [creating the release](https://docs.github.com/en/rest/releases/releases?apiVersion=2022-11-28#create-a-release) as a draft, so that we need only edit it a bit? It's a nice touch with one curl invocation :) _Originally posted by @mdbenito...
This [paper](https://arxiv.org/abs/2302.11431) makes some minor improvements to group testing. They note themselves that this is still not better than permutation sampling, so implementing the improvements would only make some sense...
In order to stop computation in workers once the overall computation is done we must signal the worker (we had this before with queues) - For joblib backend: use files...
To complete all methods in Fisher's paper