Background

The aim of this special issue is to concentrate on Performance Engineering of Artificial Intelligent Systems to improve the performance of systems that operate on big data. In particular, among artificial intelligence systems, deep learning systems, which are actively researched on these days, consume a lot of computing power and time because they continuously adjust weights in their models based on data. To this end, various optimization functions have been proposed to quickly optimize weights, but these functions provide different results according to the characteristics of data. That is, optimization functions partially contribute to the performance of deep learning systems but do not ultimately improve the performance of the systems at the system level.

In particular, we are interested in the performance of deep learning systems working on big data that continuously accumulate over time. In this case, the systems continuously adjust the weights in their models as data accumulates. We are interested in improving evaluation methods as well. The most commonly used evaluation method of deep learning techniques is k-fold cross validation, which divides the data into k groups and tests each group by training a model with other groups. A more realistic evaluation method is an online learning evaluation method, which evaluates models as data accumulates. However, deep learning techniques consume a lot of time in updating weights in deep learning models. Thus, when applying an online learning evaluation method, it takes exponentially longer to create and update models in iterative ways. In addition, deep learning techniques where online learning evaluation methods could be applied to continuously accumulating data are limited.

Accordingly, we solicit manuscripts on effective mechanisms and evaluation methods of deep learning systems that work on continuously accumulating data. In contrast, we do not solicit manuscripts on dividing data through distributed systems, processing data, and collecting the results, because these approaches only divide the same computing across the multiple systems and the total amount of computation remains the same. This special issue seeks mechanisms to reduce the amount of computation while maintaining reasonable accuracy in deep learning systems, gradual updates of the weights in deep learning systems according to the accumulation of big data over time, and online learning evaluation methods that effectively work on the continuously accumulating data.

This special issue, which focuses on the performance engineering of artificial intelligence systems, will share how to improve the performance of artificial intelligence systems for all relevant research and industry groups. It will urge a fundamental solution to the effectiveness of deep learning systems that train large data with intensive computing resources.

Topics of Interest

Researchers and practitioners may submit their innovations and findings while developing deep learning systems. Sample topics are given below:

Submission

We are soliciting original contributions that have not been published and are not currently under consideration elsewhere. Both theoretical studies and state-of-the-art practical applications are welcome. All submitted papers will be peer-reviewed and selected on the basis of their quality and relevance to the theme of this special section.

We also encourage extensions of conference papers, unless prohibited by copyright, if there is a significant difference in the technical content. Improvements such as adding a new case study or including a description of additional related studies do not satisfy this requirement. A description explaining the difference between the conference paper and the journal submission is required. The overlap between each submission and other articles, including the authors’ own papers and dissertations, should be less than 30%. Each submission must conform to the IJPE template. Please click here to submit your paper.

Special Attention

Important Dates

Guest Editors

About the Guest Editors

Professor Seonah Lee is an Associate Professor at the Department of Aerospace and Software Engineering as well as the Department of AI Convergence Engineering at Gyeongsang National University, Republic of Korea. She worked as a software engineer at Samsung Electronics from 1999 to 2006. She received her Ph.D. in Software Engineering and Recommendation Systems from the School of Computer Science, KAIST, in 2013. She has published more than 60 papers in international journals, international conference proceedings, and national conferences. She is a member of IEEE, KIISE, and KIPS. She is an editorial board member and reviewer of international and national journals and conferences. Her current research interest includes software engineering for artificial intelligent systems and artificial intelligent systems for software engineering in the perspective of software evolution.

Professor Kyong Hoon Kim is currently a professor of the School of Computer Science and Engineering at Kyugnpook National University, Korea. Before that, he was a professor in the Department of Informatics at Gyeongsang National University, Korea from September 2007 to February 2020, and a post-doctorial researcher at the University of Melbourne, Australia from September 2005 to August 2007. Professor Kim has published more than 100 papers in international journals and conference proceedings. His research interests include real-time systems, cloud computing, computer security, avionics systems, and artificial intelligence in medical applications.