Ying Wu College of Computing Professor Presents Research at Multiple Top Scientific Events in Asia
Professor Cristian Borcea from the Department of Computer Science recently returned from a brief tour of China and Japan where he presented research papers on his continuing work with federated learning for mobile devices as well as a novel approach for improving the revenue of online publishers in real-time Ad auctions. His paper on Ad auctions, done in collaboration with Forbes, was among 93 accepted out of over 1,000 submissions that were selected for presentation at the premiere research conference on Data Mining, the 23rd IEEE International Conference on Data Mining (ICDM 2023), in Shanghai, China.
Borcea’s research on Reserve Price Optimization in First-Price Auctions via Multi-Task Learning, which he conducted with Professor Yi Chen, his frequent collaborator in the Martin Tuchman School of Management, creates a machine learning model that determines optimal reserve prices for publishers selling ad impressions through first-price Ad auctions in real-time bidding, such as those managed by Google. Setting an appropriate reserve price for an ad impression is critical for publisher revenue, yet challenging, given that publishers do not know the bidding prices for failed auctions. To solve this problem, Borcea’s team proposed a multi-task learning model that learns at the same time the optimal reserve price and the failure probability of the auction. The results using data from Forbes have demonstrated success in increasing the revenue for online publishers by setting the optimal reserve price for auctions in mere milliseconds.
While in Shanghai, Borcea also went to Shanghai Normal University to deliver his presentation on Federated Learning (FL) for mobile and IoT devices at the invitation of Professor Chang Guo. Borcea and Guo have been collaborating since the 2018-2019 academic year, when Guo attended NJIT as a Ph.D. exchange student. In the past, they published research on minimizing vehicular traffic congestion using mobile-cloud systems, and now they are exploring FL techniques for the same problem.
Borcea’s last stop in his Asian trip was at the National Institute of Informatics (NII) in Tokyo, Japan. NJIT currently has an MOU (Memorandum of Understanding) with NII for research collaboration, and Borcea holds a visiting professor appointment at NII since 2012. At NII, he presented his work on FL for mobile devices to an expanded audience from the broader Tokyo area, including three FL experts from Toyota. Since Toyota is exploring options for deploying FL in vehicular systems, their experts discussed with Borcea about his FL systems and the possible collaboration toward turning these systems into open source software to achieve wide spread deployment of FL.
“Partnerships with researchers in countries such as China and Japan allow faculty and students to exchange ideas and experience different perspectives,” said Borcea. “For example, Japan favors a very hands-on approach to research. This aligns extremely well with the practice-based education, training and research we conduct at NJIT, and we have a vested interest in fostering this bourgeoning relationship with such countries as an important part of building a global tech ecosystem.”