Jonas Landsgesell comments

Results 17 comments of


                                            Jonas Landsgesell

Monte Carlo displacement move is broken

Hi Jean-Noel The function implements a multi-particle MC move. It moves 1...N distinct particles (by design) where N is defined by the calling function. Good that you catched point 4....

[Question] Is auto-sklearn ensemble for classification using predict_proba + threshold adaptions

Nice to read this open discussion and seeing benchmarks for the idea. @LennartPurucker picking up your question on how to deal with model dependent thresholds I had the following thoughts....

[Question] Is auto-sklearn ensemble for classification using predict_proba + threshold adaptions

The pull request https://github.com/scikit-learn/scikit-learn/pull/16525 might implement a calibration method for tuning thresholds in sklearn. Would that be applicable for usage in Auto-Sklearn?

UMAP produces different results with the same parameters for a loop

Have a look at https://github.com/rapidsai/cuml/issues/5099#issuecomment-1396363524 Providing a random state and init="random" to UMAP (i.e. UMAP(n_neighbors=15, random_state=1, init="random")) generates deterministic behavior for me.

Update _bertopic.py to fix question/ github issue #1696

Sure! Do you have a suggestion for a specific wording? I am currently lacking the fantasy for other ways to express the fact that HDBSCAN is responsible here while we...

topic_model.transform(docs)[0][i] is sometimes different from topic_model.transform(docs[i])[0][0]

Thank you for your clarification! Is it worth adding a note to the documentation to prevent future questions?

LearningShapelets implmentation for imbalanced dataset in the params providing class_weights and loss will be helpful

A differentiable version of the score as a loss function during training would also be interesting for imbalanced classes in a binary classification setting: https://stackoverflow.com/a/65320239/12229416