9.2. Problems of imbalance in graph data
135
(a) Synthetic — two classes (250, 500, 750, and 1000 nodes).
(b) USPS — two classes (250, 500, 750, and 1000 nodes).
(c) ISOLET — two classes (75, 150, 225, and 300 nodes).
Figure 9.9: Error rates for four classes with the same number of labelled nodes but
different numbers of nodes in each class
135
(a) Synthetic — two classes (250, 500, 750, and 1000 nodes).
(b) USPS — two classes (250, 500, 750, and 1000 nodes).
(c) ISOLET — two classes (75, 150, 225, and 300 nodes).
Figure 9.9: Error rates for four classes with the same number of labelled nodes but
different numbers of nodes in each class
