کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
492890 | 721660 | 2014 | 9 صفحه PDF | دانلود رایگان |

We consider an active supervised learning scenario in which the supervisor (trainer) can make decisions regarding the possibility to choose new examples for learning. In the classical forms of supervised learning, the training set is chosen according to some known or random given distribution. The supervisor is a passive agent in the sense that he is not able to interact with the training set in order to improve the performances of the learning process of the neural network. We will introduce in a formal manner the terms of “difficult learning” and “easy learning” related to the training data set. We will investigate some possibilities that allow the trainer to become active and we will analyse the performances of such supervised learning. An active supervised learning algorithm is presented and also we have performed some simulation in order to prove our theoretical results.
Journal: Procedia Technology - Volume 12, 2014, Pages 220-228