TRR318 / scikit-psl

Probabilistic Scoring System – a probabilistic & incremental extension to Scoring Systems

TRR318/scikit-psl Issues

Integrate beta calibration into our core repository
Closed 4 months ago
Validate PSL with sklearn check estimator
Updated 4 months ago
Feature Dropout during psl learning
Updated 5 months ago
Change default Stage of PSL when predicting (currently last stage is used regardless of performance)
Updated 5 months ago
Train PSL via Rank loss
Closed 5 months ago1
Representation as decision tree
Updated 6 months ago
Make Classifier robust to using pandas dataframes
Closed 6 months ago
Lookahead should respect predefined scores even if the solution is suboptimal
Updated 8 months ago
investigate crash in inspect method. seems to be related to binary datasets
Closed 8 months ago2
Fit-method should be able to accept weights for datapoints
Closed 6 months ago
implement expected entropy as function of y_true and y_proba instead of X
Closed 8 months ago
Try extracting internal Threshhold optimizer to Transformer
Updated 9 months ago
Use cachetools to cache already evaluated models during lookahead
Closed 9 months ago3
use "permutations" instead of "combinations" for lookahead and feature sequence generation
Closed 9 months ago
Add global-loss for optimizing the psl
Closed 9 months ago
merge/consolidate scoringsystem-cascadizer into the repository
Updated 9 months ago
Additional Hyperparams of the classifierAtK
Closed 9 months ago
extract expected entropy as proper sklearn metric function
Closed 9 months ago1
move hyperparameters to constructor
Closed 9 months ago
[optional] optimization of internal threshold calculation
Closed 10 months ago3
One-shot Optimization
Closed 10 months ago1
Improve binarizer to use explicit linear programs instead of Blackbox optimizer
Closed a year ago1
Create example for binarizer used in pipeline
Closed a year ago1
Extend PSL to be capable of working with continuos datasets
Closed a year ago2
implement psl.describe(·)
Closed a year ago
Subset summation in Classifier at k is wrong
Closed a year ago1
first negative correlated feature gets minimal negative score instead of maximal
Closed a year ago