Special issue on Adaptive Machine Learning in the IoT-Edge-Cloud Continuum

Emerging availability (and varying complexity and types) of Internet of Things (IoT) devices, along with large data volumes that such devices (can potentially) generate, can have a significant impact on our lives, fueling the development of critical next-generation services and applications in a variety of application domains (e.g. healthcare, smart grids, finance, disaster management, agriculture, transportation and water management, etc). Market forces also fuel the significant increase in the types of IoT devices, as many vendors are now embedding network capability with support for standard protocols (e.g. HTTP/REST) within their products, which were initially stand-alone. Understanding how data from such devices can be more efficiently analysed using Machine Learning (ML) techniques remains a challenge, with existing reliance on large scale Cloud computing systems becoming a bottleneck over time. Transferring large data streams to such centralised Cloud data-centre environments, in a timely and reliable manner, is being seen as a key limitation of current technologies.

Many researchers have investigated ML-based techniques to analyse IoT data in the Cloud datacenters. However, classical cloud-hosted ML techniques focus on single central node with full access to global training dataset. As massive amount of storage and computing power is needed for training the ML classifiers, existing approaches are not suitable for resource constraint IoT and edge devices, which simply do not have the memory or processing capability to train and sometimes even execute the models.

Progressing beyond state-of-the-art new ML techniques need to be investigated that could be made use of across a heterogeneous pool of computational resources in the IoT-Edge-Cloud continuum. We believe the combined use of edge devices and large-scale data-centres could lead to a new class of ML techniques that are "resource aware", and able to take account of a resource on which the algorithm is executed -- trading accuracy with execution time.

This proposal aims to collect several articles covering algorithms, methods, and computing frameworks for adaptive machine learning in connection with IoT-Edge-Cloud based applications in healthcare, smart grids, finance, disaster management, agriculture, transportation and water management. 

The topics of interest for this special issue include, but are not limited to:

  • Benchmarking ML kernels for resource constraint devices
  • Performance models for optimizing ML techniques at the extreme edge
  • Performance monitoring techniques for federated ML
  • Efficient networking and communication techniques for federated ML
  • Security and privacy for federated ML
  • Techniques for handling IoT data uncertainties and incompleteness
  • Hardware/software co-design approaches
  • Application case studies

Guest Editors

Prof. Rajiv Ranjan (Lead Guest Editor), School of Computing, Newcastle University, 1 Science Square, Newcastle Helix, Newcastle upon Tyne, UK, NE4 5TG, Email: raj.ranjan@ncl.ac.uk
Prof. Dan Chen, School of Computer Science, Wuhan University, Bayi Road 299#, Wuhan, China, 430072, Email: dan.chen@whu.edu.cn
Associate Professor Prem Prakash Jayaraman, School of Software and Electrical Engineering, Swinburne University of Technology, Hawthorn, VIC, Australia, Email: pjayaraman@swin.edu.au

Important Dates

Closed for submissions: November 1, 2021  
Results of first round of reviews: February 1, 2022  
Submission of revised manuscripts: May 1, 2022
Results of second round of reviews: July 1, 2022

Major Guidelines

The special issue invites original research papers that make significant contributions to the state-of-the-art algorithms, methods, and computing frameworks for adaptive machine learning in connection with IoT-Edge-Cloud based applications.
We seek submission of papers that present new, original and innovative ideas for the "first" time in Computing journal. Submission of "extended versions" of already published works (e.g., conference/workshop papers) is not encouraged unless they contain a significant number of "new and original" ideas/contributions along with more than 50% brand "new" material. Otherwise, the submission will be "desk" rejected without being reviewed. Each submitted paper will receive at least two reviews. The editorial review committee will include well-known experts in the areas of Computational Intelligence, Machine Learning, Data Science, Big Data, and Edge Computing.

Selection and Evaluation Criteria:

- Significance to the readership of the journal
- Relevance to the special issue
- Originality of idea, technical contribution, and significance of the presented results
- Quality, clarity, and readability of the written text
- Quality of references and related work
- Quality of research hypothesis, assertions, and conclusion

Author Submission Guidelines

All submitted manuscripts must be formatted according to the Computing's instructions for authors which are available at https://www.springer.com/607. We will only accept LaTeX manuscripts (which must use Springer templates at https://www/springer.com/607).

Submission instruction
Manuscripts should be submitted using the online submission system at (https://www.springer.com/607).

Authors should select ‘SI:  Adaptive Machine Learning in the IoT-Edge-Cloud Continuum’ during the submission step 'Additional Information'.