all AI news
In Defense of Softmax Parametrization for Calibrated and Consistent Learning to Defer. (arXiv:2311.01106v1 [cs.LG])
cs.LG updates on arXiv.org arxiv.org
Enabling machine learning classifiers to defer their decision to a downstream
expert when the expert is more accurate will ensure improved safety and
performance. This objective can be achieved with the learning-to-defer
framework which aims to jointly learn how to classify and how to defer to the
expert. In recent studies, it has been theoretically shown that popular
estimators for learning to defer parameterized with softmax provide unbounded
estimates for the likelihood of deferring which makes them uncalibrated.
However, it …
arxiv classifiers consistent decision defense enabling expert framework learn machine machine learning performance safety softmax