skada.ClassRegularizerOTMapping
- skada.ClassRegularizerOTMapping(base_estimator=SVC(), metric='sqeuclidean', norm='lpl1', max_iter=10, max_inner_iter=200, reg_e=1.0, reg_cl=0.1, tol=1e-08)[source]
ClassRegularizedOTMapping pipeline with adapter and estimator.
see [6] for details.
- Parameters:
- base_estimatorobject, optional (default=SVC(kernel="rbf"))
The base estimator to fit on the target dataset.
- reg_efloat, default=1
Entropic regularization parameter.
- reg_clfloat, default=0.1
Class regularization parameter.
- normstr, default="lpl1"
Norm use for the regularizer of the class labels. If "lpl1", use the lp l1 norm. If "l1l2", use the l1 l2 norm.
- metricstr, optional (default="sqeuclidean")
The ground metric for the Wasserstein problem
- max_iterint, float, optional (default=10)
The minimum number of iteration before stopping the optimization algorithm if it has not converged
- max_inner_iterint, float, optional (default=200)
The number of iteration in the inner loop
- tolfloat, optional (default=10e-9)
Stop threshold on error (inner sinkhorn solver) (>0)
- Returns:
- pipelinePipeline
Pipeline containing ClassRegularizerOTMapping adapter and base estimator.
References
[6]N. Courty, R. Flamary, D. Tuia and A. Rakotomamonjy, Optimal Transport for Domain Adaptation, in IEEE Transactions on Pattern Analysis and Machine Intelligence
