Transfer Learning for Feature Dimensionality Reduction

The international arab journal of information technology

article

Nikhila Thribhuvan¹  Sudheep Elayidom² 

[1] Department of Information Technology, Rajagiri School of Engineering and Technology;Division of Computer Science, School of Engineering, Cochin University of Science and Technology

关键词: Dimensionality reduction;    fine-tuning;    transfer learning;    VGG-16;    VGG-19;   

DOI  :  10.34028/iajit/19/5/3

学科分类：计算机科学（综合）

来源: Zarqa University

【 摘 要 】

Transfer learning is a machine learning methodology by which a model developed for achieving a task is exploitedfor another related job. Many pre-trained image classification models trained on ImageNet are used for transfer learning. Thesepre-trained networks could also be used for classifying out of domain images by retraining them. This paper, along with theexisting application for these pre-trained models, is also being exploited for feature dimensionality reduction. Manydimensionality reduction methods are available; the pre-trained image models will help us perform both image feature extractionand dimensionality reduction in a single go using the same network. The fine-tuning of the fully connected layers of the pre- trained network is done to extract the image features; along with this fine-tuning, some more tweaking is done on the fullyconnected layers of these models to reduce the image feature dimensionality. Here, VGG-16 and VGG-19 are the pre-trainedmodels considered for feature vector generation and dimensionality reduction. An analysis of the efficiency of features generatedby these pre-trained networks in classifying the out-of-domain images is done. Three different variants of VGG-16 and VGG-19are analysed. All the three variants developed gave an AUC value above 0.8, which is considered good.

【 授权许可】

Unknown