Efficient HPC libraries often expose multiple tunable parameters, algorithmic implementations, or a combination of them, to provide optimized routines. The optimal parameters and algorithmic choices may depend on input properties such as the shapes of the matrices involved in the operation. Traditionally, these parameters are manually tuned or set by auto-tuners. In emerging applications such as deep learning, this approach is not effective across the wide range of inputs and architectures used in practice. In this work, we analyze different machine learning techniques and predictive models to accelerate the convolution operator and GEMM. Moreover, we address the problem of dataset generation, and we study the performance, accuracy, and generalization ability of the models. Our insights allow us to improve the performance of computationally expensive deep learning primitives on high-end GPUs as well as low-power embedded GPU architectures on three different libraries. Experimental results show significant improvement in the target applications from 50% up to 300% compared to auto-tuned and high-optimized vendor-based heuristics by using simple decision tree- and MLP-based models.

The work described in this paper attempts to contribute to one of the most stimulating and promising sectors in the field of emotion recognition, which is health care. Multidisciplinary studies in artificial intelligence, augmented reality, and psychology stressed out the importance of emotions in communication and awareness. The intent is the recognition of human emotions, processing images streamed in real-time from a mobile device. The proposed techniques involve the use of open source libraries of visual recognition and machine learning approaches based on convolutional neural networks (CNN).

This issue features a special issue on ‘Computational Science and Its Applications’. Computational Science is the main pillar of most of the present research, industrial and commercial activities and plays a unique role in exploiting ICT innovative technologies. Owing to the latest development and the availability of high-performance computing, including parallel computing, grid computing, and cloud computing, there is a critical need to employ efficient and effective computational methods and algorithms in various applications, including computational biology, computational geometry, computational physics, computation chemistry, computational finance, graphics and visualization, scientific data management, data mining, etc. This issue features selected papers from the International Conference on Computational Science and Its Applications (ICCSA2009) held in 29 June–1 July, 2009, Kyung Hee University, Suwon, South Korea. In addition to extended papers from ICCSA2009, a special issue CFP has been distributed to a wider community through various mailing lists. Finally, we selected five papers to be included in this issue. The first paper discusses data and knowledge grids. It basically combines grid computing and real-time service management and execution paradigms, which makes use of the service-oriented architecture and paradigm suited for the grid platform. The second focuses on grid and P2P systems, especially in the context of sharing paradigm. It discusses middleware and libraries for grid and P2P systems. The third paper also focuses on P2P, whereby it describes scalable group communication protocols. The fourth paper focuses on road network query processing, which can be adopted in a mobile environment. It particularly concentrates on range search in a continuous mobile dynamic. Finally, the fifth paper introduces context-aware semantic network similarity model, using an ontological approach. As general co-chairs and program co-chairs of ICCSA2009, as well as the guest editors of the special issue on Computational Science in the Concurrency and Computation journal, we would like to congratulate the authors whose papers appeared in this special issue. We would also like to thank the PC members of ICCSA2009 who conducted the initial reviewing process for the conference and external reviewers who conducted further reviews of extended papers submitted to this special issue.