Achieving black box vertical federated learning

Wednesday, March 02, 2022

When you hear the term “black box,” most people think about the flight data recorder of an airplane. Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) Assistant Professor of Machine Learning, Dr. Bin Gu, thinks about algorithms and optimization. In computer science, black-box optimization (BBO) considers the design and analysis of algorithms for problems where the structure of the objective function and/or the constraints defining the set is unknown, un-exploitable, or non-existent (Science Direct).

Large-scale optimization

Gu has a special interest in exploring large-scale optimization techniques on bi-level optimization and black-box optimization problems to solve some of the most important machine learning tasks—automated machine learning, reinforcement learning and spiking neural networks.

“Where bilevel optimization is the fundamental mathematical problem to achieve automatic machine learning, black box optimization is especially important to machine learning due to the privacy of learning models,” Gu said.

Alongside my students, we hope to bring machine automation closer to the way the human brain operates.

Dr. Bin Gu
MBZUAI Assistant Professor of Machine Learning
Black-box optimization is a rapidly growing field of optimization and critical in many sectors such as energy, the environment, systems engineering, computational biology, drug discovery, and chemical synthesis. And Gu’s work has had some success. “We have already connected black box optimization and vertical federated learning, and achieved black box vertical federated learning which can keep data and model privacy as much as possible,” Gu continued. “This work was collaborated with JD.com – the second largest e-commerce company in China.”

Large-scale machine learning

Gu received his B.S. and Ph.D. degrees in computer science from the Nanjing University of Aeronautics and Astronautics, in 2005 and 2011, respectively. He worked as a postdoctoral fellow at the University of Western Ontario; at the University of Texas at Arlington; and at the University of Pittsburgh.

“Alongside my students, we hope to bring machine automation closer to the way the human brain operates. We have two current projects. One is large scale optimization in machine learning, and the other one is spiking neural networks.”

Gu joined MBZUAI in 2020, having previously worked as a full professor at Nanjing University of Information Science and Technology in China. Gu has published more than 70 papers, with 3000-plus citations. He served as a program committee member or reviewer for several leading machine learning and data mining conferences and journals such as NIPS, ICML, KDD, AAAI, TPAMI, JMLR, and a senior program committee member of IJCAI 2019-2021.

“In machine learning, the next step in my mind is understanding neural networks, designing human-like brain neural networks, and achieving automatic large scale machine learning,” Gu said.

Related

thumbnail
Thursday, July 17, 2025

Challenging the promise of invisible ink in the era of large models

MBZUAI's Nils Lukas and Toluwani Samuel Aremu presented research at ICML that highlights the need for more robust digital.....

  1. icml ,
  2. machine learning ,
  3. research ,
  4. conference ,
  5. ML ,
  6. paper ,
  7. watermarking ,
  8. watermark ,
Read More
thumbnail
Tuesday, July 15, 2025

New multimodal model brings pixel-level precision to satellite imagery

MBZUAI's Akashah Shabbir presented GeoPixel at ICML: a multimodal model that supports the task of pixel grounding.....

  1. geopixel ,
  2. remote sensing ,
  3. multimodal ,
  4. Ph.D. ,
  5. dataset ,
  6. icml ,
  7. computer vision ,
  8. research ,
  9. machine learning ,
Read More
thumbnail
Friday, July 11, 2025

Not your typical digital twins: How two Emirati brothers are using AI to make the UAE’s roads safer

Identical twins Abdulla and Abdulrahman Almarzooqi share more than looks. Together, the MBZUAI master’s graduates are working.....

  1. MLLMs ,
  2. commencement ,
  3. commencement 2025 ,
  4. graduates ,
  5. alumni ,
  6. graduation ,
  7. machine learning ,
Read More