Robust Representation for Data Analytics by Sheng Li & Yun Fu

Author:Sheng Li & Yun Fu
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

6.4.1 Datasets and Settings

Auxiliary Domain Dataset. Following [44], we randomly select 5000 unlabeled images from the LabelMe website2 to construct the sample set in auxiliary domain. Currently, the LabelMe dataset contains more than 100 thousand images collected from various resources, which provide us a great auxiliary domain for self-taught learning. Figure 6.2a shows some images in the LabelMe dataset.

Fig. 6.2Sample images in auxiliary domain (above) and target domain (below)

To evaluate how the data size in the auxiliary domain affects the performance of learning tasks in the target domain, we alter the number of auxiliary samples from 1000 to 5000, and compare the performance in different settings. In our experiments, we find that increasing the size of auxiliary sample set would improve the performance of learning tasks, but the improvements are marginal when the size is over 3000. Due to the space limit, we only report the results of self-taught learning algorithms under two settings that use 1000 and 3000 auxiliary images, respectively.

Target Domain Datasets. To extensively testify our approach and related methods, we utilize the following five benchmark datasets. MSRC-v1 dataset3 contains 240 images of 9 classes. Following [14], we choose 7 classes including airplane, bicycle, building, car, cow, face, tree, and each class has 30 images. This dataset owns obvious clutter and variability in the appearances of objects. The MSRC-v2 dataset is an extension of MSRC-v1. It contains 591 images of 23 object classes. Figure 6.2c shows some images in the MSRC-v1 dataset.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.