Particle swarm optimization based parameter selection technique for unsupervised discriminant analysis in transfer learn
- PDF / 991,424 Bytes
- 19 Pages / 595.224 x 790.955 pts Page_size
- 105 Downloads / 187 Views
Particle swarm optimization based parameter selection technique for unsupervised discriminant analysis in transfer learning framework Rakesh Kumar Sanodiya1 · Jimson Mathew1 · Sriparna Saha1 · Piyush Tripathi2
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The purpose of transfer learning is to utilize the knowledge gained from the existing (source) domain to enhance the performance on a distinct but related (target) domain. Existing works on transfer learning are not capable of optimizing different quality measures (components) such as minimizing the marginal distribution, minimizing the conditional distribution, maximizing the target domain variance, modeling the manifold by utilizing the common geometric properties in the source as well as the target domain at the same time. Moreover, existing transfer learning methods use conventional approaches to determine the appropriate values of their parameters, which is very hectic and time-consuming. Therefore, in order to overcome the drawbacks of existing approaches, we propose a Particle Swarm Optimization based Parameter Selection Approach for Unsupervised Discriminant Analysis (UDATL-PSO) in transfer learning framework. In UDATLPSO, all the quality measures are considered at the same time, as well as the PSO approach has been used to select the best values of their parameters. Extensive experiments on various transfer learning tasks show that the proposed method has a significant influence on state-of-the-art methods. Keywords Unsupervised discriminant analysis · Transfer learning · Particle swarm optimization · Domain adaptation · Classification · Parameter selection
1 Introduction The rapid online application development for content sharing, multimedia, and online business has created a strong need for automated image recognition, analysis of multimedia, and other data [1, 2]. Nowadays, machine learning approaches, specially supervised learning, have Rakesh Kumar Sanodiya
[email protected] Jimson Mathew [email protected] Sriparna Saha [email protected] Piyush Tripathi [email protected] 2
Department of Computer Science, Engineering, Indian Institute of Technology Patna, Patna, India
2
IIEST Shibpur, West Bengal, India
been widely used for the development of such recognition and analysis systems [3, 4]. However, these supervised learning approaches require large amount of labeled information and the process of acquiring labeled data is often time-consuming and hectic. Therefore, it is necessary to use labeled samples in many existing related source domains to expedite learning in the new target domain [5, 6]. Conventional machine learning techniques require that the training and test domains must follow the same data distributions. However, in the real world, the distributions are different. Hence, the traditional or conventional machine learning approaches [7, 8] do not work in real-world situations. For example, Fig. 1 displays six different images of two subjects, Motorbike and Cat. Each of the images ha
Data Loading...
