Prof. Mingyi Hong Receives Intel-NSF Research Award and 2020 IBM University Award

Prof. Mingyi Hong was recently named the recipient of 2 awards: an Intel-NSF joint research award to develop artificial neural networks for wireless network systems, and the 2020 IBM University Award.

INTEL-NSF JOINT RESEARCH AWARD

The Intel-NSF joint research award for wireless communication stems from the program, Machine Learning for Wireless Network Systems (MLWiNS), the latest effort in a series of programs jointly undertaken by Intel and NSF to encourage and speed up innovations in wireless communications for future applications. 

The MLWiNS program lays particular emphasis on machine learning as it has the potential to manage and support the growing complexity and density of a burgeoning wireless network.

With this award, the University of Minnesota Twin Cities will partner with Northwestern University, and Oregon State University to work on a project titled, “Artificial Neural Networks for Interference Limited Wireless Networks.” 

The goal of the project is to develop solutions for wireless access networks to meet future demand for wireless data services by increasing capacity and maximizing network utilization while also overcoming the limitations of current solutions. Under the aegis of the Intel-NSF grant, the project will introduce novel artificial neural networks (ANN) based tools to enable a new data-driven design of next generation wireless systems.

The groups’ research will entail merging supervised and unsupervised learning techniques with time-tested models of physical resources, channels, traffic, and network utilities. This involves the critical task of exploiting commonalities of ANN-based solutions for a number of subproblems that will then contribute to the development of solutions for the larger wireless networking problem. The driving questions that Mingyi and researchers at the partner institutions will seek to answer are when, how, and why ANN-based learning techniques can be applied to a wide range of wireless networking problems with realistic constraints. The intent is to develop scalable, efficient, and adaptive solutions supported by theoretical backing that together can offer generalizable design principles. These ANN-based solutions are expected to be a major building block for next generation wireless access networks. The project is aimed to benefit industry, as well as academic research. According to Mingyi, “The framework that will be designed in this project will enable the researchers to properly integrate classical model-based approaches with the emerging data-driven approaches, and will have the potential of finding new algorithms, and wireless system architectures that are far more efficient than the state-of-the-art.”

Commenting on the significance of his work, Mingyi says: We envision that data-driven approaches, such as the ANN based learning techniques to be designed in this project, will be fundamental to the design of next generation wireless networks.

THE NEED TO INNOVATE WIRELESS COMMUNICATIONS

There is a pressing need to explore and introduce innovations in wireless communications that can be available to vertical and horizontal markets and consumers. The Intel press release on the awards spells it out explicitly: “As demand for advanced connected services and devices grows, future wireless networks will need to meet the challenging density, latency, throughput and security requirements these applications will require. Machine learning shows great potential to manage the size and complexity of such networks – addressing the demand for capacity and coverage while maintaining the stringent and diverse quality of service expected from network users. At the same time, sophisticated networks and devices create an opportunity for machine learning services and computation to be deployed closer to where the data is generated, which alleviates bandwidth, privacy, latency and scalability concerns to move data to the cloud.”

As we all log on and connect to the internet, and as the number of such devices connecting increases exponentially, we have to turn our attention to how and whether our existing networks can support such device density. (Estimates indicate that by 2025, approximately 75 billion devices will be connected to the internet.)

In a conversation with Forbes contributor John Koetsier, Thyaga Nandagopal, a deputy director with the NSF, says: “Traditional 4G networks that your current mobile devices rely on typically can support a region that has about 300 to 2,000 devices in their coverage area. We are thinking about device densities [with] tens of thousands in a small region … [going] all the way up to millions of devices in a coverage area of a single cell site in a wireless network.”

MLWiNS, the Intel-NSF joint program, was established to develop improvements to wireless networking in such extremely device dense environments. A highly competitive program, it has granted 15 awards to multiple institutions. 

More information about the awards and descriptions of projects are available in the Intel newsroom and press release fact sheet.

2020 IBM UNIVERSITY AWARD

Prof. Mingyi Hong was also recently recognized with the 2020 IBM University Award for his contributions to decentralized and adversarial learning in his project, “Decentralized Robust Adversarial Learning for Scalable Trustworthy AI.” The award promotes research and innovation in areas that comprise strategic interest for IBM as well as academic institutions.

Mingyi’s work will address a significant limitation of today’s machine learning (ML): the lack of robustness against adversarial attacks. This research project proposes a novel framework to significantly improve the efficiency in training machine learning systems so that they can be robust under such attacks. By utilizing cutting-edge decentralized optimization-based techniques, the proposed design will be able to utilize large-scale, and heterogeneous distributed computational resources, and obtain 100x speedup compared to the state-of-the art schemes. The proposed work is expected to have impact beyond the theoretical domain. By using the algorithms developed in this work, companies can release officially robust pre-trained ML models, and users can then fine-tune their down-stream robust ML models.

Nominations for the IBM University awards are through IBM employees who have technical and/or research collaborations with full-time faculty at an accredited institution that has a graduate program in their field. Nominees should have made outstanding contributions to their field or show great professional promise.

Prof. Mingyi Hong earned his doctoral degree in systems and information engineering from the University of Virginia in 2011. His research interests are in the areas of design and optimizing future generations of networks including wireless and energy networks; theory and methods for statistical and distributed signal processing; the theory and methods of large-scale optimization; the theory and methods for machine learning and big data.