Edge Machine Learning for Smart IoT Environments

The Internet of Things (IoT) envisions an ubiquitous communication and computing environment where sensors, actuators, smartphones, and other smart devices, will be networked together to offer better services for different vertical sectors including healthcare monitoring, automotive, energy, smart cities, etc. Huge amount of information is typically generated and/or collected by IoT devices due to, e.g., ubiquitous communication, imaging, social networking, as well as pervasive sensors on mobile phones, surveillance cameras, and drones, which collect streaming data on every bit of our lives. Mining information from such massive volumes of data promises to bring huge scientific and economical advancements, together with an improvement in the quality of our lives. Also, if we consider the fourth industrial revolution (a.k.a. Industry 4.0), embedding intelligent devices in the production system can revolutionize the way our industrial processes are managed, enabling distributed proactive sensing and control mechanisms aimed at preventing performance degradation and optimizing the overall production chain. The vision is for ubiquitous smart network devices to enable data-driven optimization and learning algorithms for distributed and online network operation and management, adaptable to the dynamically evolving network landscape with minimal need for human intervention.

Leveraging advances in embedded systems, contemporary IoT devices are featured with small-size and low-power designs, but their computation and communication capabilities are limited. At the same time, classical machine learning (ML) algorithms are severely demanding in terms of energy, memory and computing resources, limiting their adoption for resource constrained IoT devices. A prevalent solution during the past decade was to move computing, control, and storage resources to the remote cloud. Yet, the cloud-based IoT architecture is challenged by high latency due to direct communications with the cloud, which certainly prevents real-time applications such as, e.g., augmented reality or self-driving vehicles, which cannot afford latency, and must operate under high reliability, even when network connectivity is lost. Thus, the new breed of intelligent devices and high-stake applications (drones, augmented/virtual reality, autonomous systems, etc.), requires a novel paradigm change calling for distributed, low-latency and reliable ML at the wireless network edge, also called edge machine learning . A key technology enabler for edge ML is Mobile Edge Computing (MEC), which brings cloud computing functionalities at the edge of the network, allowing the offloading of sophisticated applications from IoT devices to small data centers, called Mobile Edge Hosts (MEH), which are located at the Radio Access Point (RAP), or at an aggregation point of the core network, thus guaranteeing low latency services and high energy efficiency. However, the latency and reliability of edge ML have to be examined with respect not only to communication but also to decentralized ML training and inference processes, in a joint and holistic manner. Interestingly, the design of this new class of learning systems opens up unprecedented possibilities for architecture modelling, analysis, and optimization at all levels of the network. Achieving this goal poses a fundamental big challenge, which represents the context of this workshop proposal.

The aim of this workshop is to collect from academic and industrial players papers reporting original, previously unpublished research, which addresses this important field. The research being presented can come from any topic area relevant to edge machine learning for IoT environments including, but not limited to, the following topic areas:

  • Machine learning at the network edge
  • Fog and edge computing for IoT
  • Resource allocation techniques for edge computing and machine learning;
  • Cooperative and distributed learning for smart environments;
  • Applications to Internet of Things, Industry 4.0, smart cities, intelligent transportation, etc.

Important Dates

Paper submission: August 22, 2019 September 6, 2019 (extended)

Notification: September 25, 2019

Camera Ready: September 30, 2019

Author Registration: September 30, 2019

Workshop Day: November 13, 2019

Submission Instructions

Paper submissions will be peer-reviewed.

The length of a paper submission including figures and references may not exceed 10 pages. Contributions must be written in English and submitted in PDF format. For preparation of papers please follow the instructions for authors for preparing plain proceedings papers; with page numbers; available at the CEUR Web page (http://ceur-ws.org/Vol-XXX/samplestyles/).

At least one of the accepted paper’s authors should attend the conference to present the work.

Authors are welcome to submit using the conference review system: https://easychair.org/conferences/?conf=edging19

Workshop Proceedings

Accepted and revised papers will be published as an online CEUR Volume, indexed by SCOPUS. Submissions should follow the publication policies of CEUR, especially paying attention to format and copyright requirements detailed here.

Workshop Chairs

Dr. Emilio Calvanese Strinati (CEA Leti, France)

Dr. Paolo Di Lorenzo (Sapienza University of Rome, Italy)

Technical Chairs

Dr. Mattia Merluzzi (Sapienza University of Rome, Italy)

Dr. Antonio De Domenico (CEA Leti, France)

Publicity Chair

Dr. Stefania Sardellitti (Sapienza University of Rome, Italy)