Abstract : This paper proposes to consider the region classification task in the context of instance-transfer learning. The proposed solution consists of the conformal algorithm that employs a nonconformity function learned by the Transfer AdaBoost algorithm. The experiments showed that our approach results in valid class regions. In addition the conditions when instance transfer can improve learning are empirically derived.
https://hal.inria.fr/hal-01459635 Contributor : Hal IfipConnect in order to contact the contributor Submitted on : Tuesday, February 7, 2017 - 1:05:52 PM Last modification on : Thursday, March 5, 2020 - 5:41:28 PM Long-term archiving on: : Monday, May 8, 2017 - 2:21:49 PM