Figure 9 - uploaded by Yinpeng Chen
Content may be subject to copyright.
Confusion matrices of four different variations: (a) baseline-1 (b) baseline-2, (c) BiC, (d) upper bound. Both baseline-1 and baseline-2 have strong bias towards new classes. BiC is capable to remove most of the bias and have similar confusion matrix with the upper bound. (Best viewed in color)
Source publication
Modern machine learning suffers from catastrophic forgetting when learning new classes incrementally. The performance dramatically degrades due to the missing data of old classes. Incremental learning methods have been proposed to retain the knowledge acquired from the old classes, by using knowledge distilling and keeping a few exemplars from the...
Contexts in source publication
Context 1
... confusion matrices of these four variations are shown in Fig. 9. Clearly, baseline-1 and baseline-2 suffer from the bias towards the new classes (strong confusions on the last 20 classes). BiC reduces the bias and has similar confusion matrix to the upper ...
Similar publications
Long instrumental records are fundamental to study climate changes, but their reliability and quality have to be checked before any statistical research. Moreover, when metadata are used to solve some problems, data interpretation may change and require further work to refine the series. A thorough revision of the three-century precipitation series...
The goal of this study was to define an empirical model to calculate the leaf area in rice from linear leaf measure in genotypes used by farmers in Brazil. Through the leaf dimensions it is possible to identify the final crop yield from the LAI. Therefore, the leaves shape is closely related to the production of photoassimilates that will be conver...