Evolutionary Programming Based Deep Learning Feature Selection and Network Construction for Visual Data Classification
- PDF / 2,043,343 Bytes
- 14 Pages / 595.224 x 790.955 pts Page_size
- 9 Downloads / 195 Views
Evolutionary Programming Based Deep Learning Feature Selection and Network Construction for Visual Data Classification Haiman Tian1
· Shu-Ching Chen1 · Mei-Ling Shyu2
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Convolutional Neural Network (CNN) models and many accessible large-scale public visual datasets have brought lots of research work to a remarkable new stage. Benefited from well-trained CNN models, small training datasets can learn comprehensive features by utilizing the preliminary features from transfer learning. However, the performance is not guaranteed when taking these features to construct a new model, as the differences always exist between the source and target domains. In this paper, we propose to build an Evolution Programming-based framework to address various challenges. This framework automates both the feature learning and model building processes. It first identifies the most valuable features from pre-trained models and then constructs a suitable model to understand the characteristic features for different tasks. Each model differs in numerous ways. Overall, the experimental results effectively reach optimal solutions, demonstrating that a time-consuming task could also be conducted by an automated process that exceeds the human ability. Keywords Deep learning · Evolutionary programming · Image classification
1 Introduction The Internet era keeps people connected through all kinds of digital devices. For example, interactions happen during the active use of mobile phones, tablets, the Internet of Things (IoT) devices, vehicles, and smart household appliances. These devices are all connected and can further affect our everyday lives by generating and conveying waves of data in one second (Chang 2019; Mukherjee 2020). Along with this data generation, multimedia data (Chen et al. 1998, 2001, 2005), as a vital part that makes up 70% of the daily Internet traffic, has been utilized frequently to solve
Haiman Tian
[email protected] Shu-Ching Chen [email protected] Mei-Ling Shyu [email protected] 1
School of Computing and Information Sciences, Florida International University, Miami, FL 33199, USA
2
Department of Electrical and Computer Engineering, University of Miami, Coral Gables, FL 33124, USA
problems at various domains for both industrial applications and academic research (Li et al. 2002; Pouyanfar et al. 2018; Zhu et al. 2011, 2015). Various advanced techniques are popularly used to take the full advantage of these multimedia data in different research fields (Chen et al. 2013; Chen and Kashyap 2001; Chen et al. 2006; Lin and Shyu 2010). Among these, Deep Learning (DL) approaches (Pouyanfar et al. 2018) have generated many astonishing research outcomes in different areas, such as multimedia research including image classification (Tian et al. 2018), speech recognition (Wang and Zheng 2015), video understanding, etc. However, DL approaches are usually time-consuming and computationally expensive. Hence, a DL model built from scratch for an individ
Data Loading...
