Subspace Learning Based on Data Distribution for Face Recognition
Over the past years, a large family of algorithms has been designed to provide different solutions to the problem of dimensionality reduction, such as discriminant neighborhood embedding (DNE), marginal fisher analysis (MFA) and double adjacency graphs-based discriminant neighborhood embedding (DAG-DNE). In this paper, we investigate the effect of data distribution for face recognition. We conduct three settings to investigate the performance when we have different numbers of the training samples. One is randomly select 20% samples as training set and the remaining face images are used for testing. One is randomly select 40% samples as training set and the last one is randomly select 60% samples as training set. In the end, we find as interesting observation is that when the training sample size is large enough to sufficiently characterize the data distribution, all algorithms we discussed in this work can achieve good performance.
KeywordsDiscriminant neighborhood embedding Marginal fisher analysis DAG-DNE
- 2.Liu, Z., Wang, S., Sun, Q., Zou, H., Yang, F.: Cost-aware cloud service request scheduling for saas providers. Mob. Inf. Syst. 57(2), 291–301 (2014)Google Scholar
- 5.Wang, S., Hsu, C.-H., Liang, Z., Sun, Q., Yang, F.: Multi-user web service selection based on multi-qos prediction. Mob. Inf. Syst. 16(1), 143–152 (2014)Google Scholar
- 8.Wang, S., Sun, Q., Zou, H., Yang, F.: Towards an accurate evaluation of quality of cloud service in service-oriented cloud computing. Mob. Inf. Syst. 25(2), 283–291 (2014)Google Scholar
- 9.Wang, S., Zheng, Z., Zhengping, W., Yang, F.: Context-aware mobile service adaptation via a co-evolution extended classifier system in mobile network environments. Mob. Inf. Syst. 10(2), 197–215 (2014)Google Scholar