Different sets used in literature for active learning.

Contexts in source publication

Context 1

... very few training instances. When classifiers try to generalize from only a few training samples, their performance might be very sensitive to small changes. Also, the performance probably varies a lot depending on the concrete choice of instances to be labeled. Hence, lots of repetitions are needed to get a reliable trend of the performance. In Fig. 2, we clarify the nomenclature of different sets that might take part in ...

Context 2

... expectations held to only a small degree; moreover for targets education as well as income there was no clear winner amongst the weight categories, with some achieving better or worse depending on a specific factor of k. We got the smoothest results for the marital status target, with human bias winning consistently over equal weights as well as human interaction ( Figure 2). We interpret this as stemming from the fact that there is a significant correlation between the attributes 'marital-status' and 'relationship' in the dataset, which led users to consciously overvalue the latter when prompted for their bias. ...

Context 3

... on glass) and a SVM with RWM kernel (+0.02 on haberman). However, these gains are accompanied by worse performance than McPAL on other data sets. The advantage provided by the foreign classifiers reduces in the stages of 20 and 30 sampled instances, with svm gmm still showing good gains at 20 sampled instances (+0.05 on steel, +0.1 on vehicle). Fig. 2 shows the learning curves on the vehicle and steel data sets. While on ve- hicle a solid advantage of svm gmm , svm rbf and pwc gmm over McPAL in terms of accuracy can be observed, the development on steel is a different one. While the svm gmm and svm rwm learners perform well due to a stagnating but better performance in the early ...

Context 4

... difference between class based experience and region based experience is showed in Figure 2. Here, we have a region of the input space where two classes strongly overlap, green •'s and blue +. ...