August 17, 2024
LIDS PhD candidate Brice Huang has received the Machtey Award for Best Student Paper at FOCS 2024 for his paper, “Capacity threshold for the Ising perceptron.” The award is given annually at the IEEE Symposium on Foundations of Computer Science (FOCS) to the most outstanding paper(s) written solely by one or more students, and will be conferred this October at the 65th IEEE Symposium on Foundations of Computer Science.
Huang’s paper deals with the Ising perceptron, a model of a neural network whose study dates back to the 1960s, and addresses the question: how much high-dimensional data can this neural network memorize? In 1989, Krauth and Mézard gave a famous conjecture for the answer, which has remained open for 35 years. In 2018, Jian Ding (Peking University) and Nike Sun (MIT Mathematics) showed the answer to this question is at least this conjectured value, making significant progress on this problem. In his paper, Huang shows the answer is also at most the conjectured value, which completes the proof of Krauth and Mézard's conjecture.
Brice Huang is a fifth-year PhD student in LIDS, jointly advised by Guy Bresler (EECS, LIDS) and Nike Sun (Mathematics). His research is on high-dimensional probability, with a focus on optimal algorithms for high-dimensional statistical tasks. He earned a SM in EECS in 2022 and a SB in Mathematics and EECS in 2019, both from MIT. He was a recipient of the Ernst A. Guillemin SM Thesis Award in Artificial Intelligence and Decision-Making, a Google PhD Fellowship, and a Siebel Scholarship.
The IEEE Annual Symposium on Foundations of Computer Science (FOCS) is an academic conference in the field of theoretical computer science. FOCS is sponsored by the IEEE Computer Society. The IEEE Computer Society traces its origins to the 1946 formation of the Subcommittee on Large-Scale Computing Devices (LCD) of the American Institute of Electrical Engineers (AIEE). Over the years, it evolved to become the Institute of Electrical and Electronics Engineers (IEEE) in 1963. The Computer Society celebrated its 75th anniversary in 2021.
Read the paper. https://arxiv.org/abs/2404.18902