With the advent of the 5G era, it greatly promotes the development of mobile networks by tremendously enhancing the transmission capacity and speed and, in addition, the popularization of the Internet and various multimedia platforms, which causes the increase of dimensions, diversity and complexity of data. It is an urgent need to explore efficient dimensionality reduction techniques for mining information from mass data. As the major research contents of dimensionality reduction, feature selection aims to obtain an optimal feature subset in the original space, conversely, feature extraction attempts to find an appropriate low-dimensionality space to represent the given data. Both approaches can dramatically reduce the storage requirements and further improve the computing performance. In the past, feature selection and extraction have achieved great success in machine learning fields such as classification, regression and clustering.

Although the current research of dimensionality reduction has shown promising results, it still faces challenges. The models of dimensionality reduction are inadequate in processing large-scale, noisy, and multisource data. The efficiency of the optimization algorithms is urgently desired to be improved. Furthermore, the emerging deep learning shows the encouraging ability to extract informative features, but the interpretability still needs to be investigated.

This Special Issue encourages original research and review articles on high performance computing models for dimensionality reduction in complex conditions. We also welcome the analysis and design of efficient optimization algorithms, as well as the interpretability of deep learning.

Potential topics include but are not limited to the following:

• Scientific programming for feature selection
• Scientific programming for feature extraction
• Scientific programming for deep feature learning
• Scientific programming for ensemble learning
• Supervised and unsupervised learning
• Graph-based clustering
• Parallelization of feature learning
• High-performance knowledge discovery for multisource data
• Optimization algorithm design and analysis
• Interpretability of deep learning