Kristian Agasøster Haaga

Results 104 issues of Kristian Agasøster Haaga

Zan et al. (2022) a new CMI estimator for mixed data and introduces the local adaptive permutation test, a variant of the `LocalPermutationTest`, which we already implement. The paper is...

new estimator
independence tests
feature request

Assaad et al. (2022) presents a new PC/FCI-like algorithm here: https://www.mdpi.com/1099-4300/24/8/1156. They also introduce a variant of the (conditional) mutual information that might be deserving of its own function.

new algorithm
association measures
feature request

The PCMCI/PCMCI+ algorithms are state-of-the-art when it comes to detection causal graphs from data. They should be included in the library. ## References - Runge, J., Nowack, P., Kretschmer, M.,...

new algorithm
feature request

Colombo et al. (2012) introduces the *really* fast causal inference (RFCI) algorithm. We should implement it. ## References Colombo, D., Maathuis, M. H., Kalisch, M., & Richardson, T. S. (2012)....

new algorithm
feature request

We should implement the [dual total correlation](https://en.wikipedia.org/wiki/Dual_total_correlation), which is a generalization of the mutual information.

association measures
feature request

[Kubkowski et al., 2021](https://www.jmlr.org/papers/volume22/19-600/19-600.pdf) paper introduces the short expansion of CMI. It also contains an associated conditional independence test. We should implement both the measure and the independence test. ##...

association measures
independence tests
feature request

[Mukherjee et al. (2020)](http://proceedings.mlr.press/v115/mukherjee20a.html) introduces some classifier-based estimators of CMI based on neural networks. It would be nice to have these at some point. However, it would require some design...

hard
design-discussion
new estimator
conditional mutual information
feature request

Marx et al. (2021) introduces an adaptive-histogram based estimator for CMI. We should implement it. ## References Marx, A., Yang, L., & van Leeuwen, M. (2021). Estimating conditional mutual information...

new estimator
conditional mutual information
feature request

Do the same as in #343 for `OCE` and `PC` algorithms with `infer_graph`

quality of life

As mentioned in #341, independence testing can be very slow. The slow runtime is mostly due to intrinsic properties of the methods. However For example, in #341 it is mentioned...

performance