Feature Selection for Big Visual Data Overview and Challenges

摘要

The unprecedented amount of visual data that is available nowadays has created new research opportunities and challenges in the areas of computer vision and machine learning. When dealing with large scale datasets, with a huge number of samples and features, the use of feature selection plays an important role for dimensionality reduction whilst allowing model interpretation, data understanding and knowledge extraction. This manuscript is focused on feature selection as applied to big visual data, including both traditional and deep approaches, and tries to give an overview of the cutting-edge techniques to deal with large-scale vision problems and identify technical challenges in the field.