Separate fitting from prediction in `ResspectClassifier`

LSSTDESC / RESSPECT

The RESSPECT project is a result from an inter-collaboration agreement established between the Cosmostatistics Initiative (COIN) and the LSST Dark Energy Science Collaboration (DESC) with the goal of developing a recommendation system for telescope resource allocation able to optimize photometric supernova cosmology anaylsis.

MIT License

1 stars 0 forks source link

Separate fitting from prediction in `ResspectClassifier` #51

Closed drewoldag closed 1 month ago

drewoldag commented 1 month ago

Streamlined RessepectClassifier base class, no need to instantiate with training and test data.
Move bootstrap outside of base class to support potential scenario of multiple types of classifiers in an ensemble.
Implemented a simple load_classifier function.
Updated database.py to match the changes in classifiers.py.
Changed PreFitVotingClassifier to work with classifier classes instead of free functions.

github-actions[bot] commented 1 month ago

Before [4e130d16]	After [c7ad61bd]	Ratio	Benchmark (Parameter)
173M	175M	1.01	benchmarks.peakmem_learn_loop('KNN')
171±1ms	173±3ms	1.01	benchmarks.time_learn_loop('KNN', 'UncSampling')
168M	168M	1	benchmarks.peakmem_learn_loop('RandomForest')
136±0.8ms	137±2ms	1	benchmarks.time_feature_creation
2.64±0.01s	2.64±0.01s	1	benchmarks.time_learn_loop('RandomForest', 'RandomSampling')
2.68±0.02s	2.67±0.02s	0.99	benchmarks.time_learn_loop('RandomForest', 'UncSampling')
169±1ms	165±0.8ms	0.98	benchmarks.time_learn_loop('KNN', 'RandomSampling')

Click here to view all benchmarks.