Closed BradKML closed 3 years ago
Hi, some CDlib algorithms require additional, mandatory, parameters for which there is not a default value. Thus, this is not a bug but an expected behavior.
For such a reason the code snippet you provided cannot be applied as a default solution to instantiate iteratively all exposed algorithms.
Moreover, there is already a way to pool algorithms. See https://cdlib.readthedocs.io/en/latest/reference/ensemble.html#multiple-instantiation
@GiulioRossetti what if instead of pooling, we are trying to do cross-model evaluations?
@GiulioRossetti what if instead of pooling, we are trying to do cross-model evaluations?
To perform cross-model evaluation/comparison, given a graph you still need to compute a set of clusterings, each applying a different algorithm to it. For such a purpose pooling (or even grid search if you aim to fine-tune model parameters) is fine since it only offers a standard way to structure multi-model experiments.
Describe the bug
To Reproduce Steps to reproduce the behavior:
Expected behavior expect that this would work since other algorithms checks out.
[algorithms.aslpaw, algorithms.chinesewhispers, algorithms.edmot, algorithms.der, algorithms.cpm, algorithms.ga, algorithms.gdmp2, algorithms.greedy_modularity, algorithms.infomap, algorithms.significance_communities, algorithms.surprise_communities, algorithms.walktrap, algorithms.sbm_dl, algorithms.sbm_dl_nested]
Utility To quickly iterate through multiple algorithms at once