Powered by Indico
v3.2.3
Host: Prof. Xin Liu
Venue: TDLI Meeting Room N400
Tencent Meeting link: https://meeting.tencent.com/dm/CcYqQPEFY8zm
Meeting ID: 491786622, no password
Abstract:
Quantum error correction is essential for bridging the large gap between the error rates of current quantum hardware and those required for practical fault-tolerant computation. A central challenge is decoding: inferring the most probable logical error class from noisy syndrome measurements. While minimum-weight decoding identifies a single low-weight error configuration, maximum-likelihood decoding accounts for the combined probability of all logically equivalent configurations and can therefore provide higher decoding accuracy, albeit at substantially greater computational cost.
This talk presents three complementary approaches to maximum-likelihood decoding. First, we show how the decoding problem for planar codes can be mapped to a statistical-mechanics model whose partition function is evaluated efficiently using the Kac–Ward formulation. Second, we discuss tensor-network representations that extend maximum-likelihood decoding to more general code structures, together with methods for accelerating approximate contraction. Finally, we examine neural-network decoders that learn correlations in syndrome histories and adapt to realistic device noise through simulation-based pretraining and experimental fine-tuning. Results on representative-code benchmarks and experimental quantum-device data illustrate the accuracy, scalability, and computational trade-offs of these approaches. Together, they provide a unified perspective on using statistical mechanics, tensor networks, and machine learning to enable practical, high-accuracy quantum error correction.
Biography:
Feng Pan is an Assistant Professor in the Science, Mathematics, and Technology Cluster at the Singapore University of Technology and Design. He received his PhD in theoretical physics from the Institute of Theoretical Physics at the Chinese Academy of Sciences in 2022. Following his PhD, he worked as a Research Fellow at the Centre for Quantum Technologies.
His research lies at the intersection of tensor networks, quantum computing, statistical physics, machine learning, and high-performance scientific computing. His work has been published in leading journals and conferences, including Nature Computational Science, Physical Review Letters, and the International Conference for High Performance Computing, Networking, Storage and Analysis. His notable contributions include algorithms for contracting arbitrary tensor networks, large-scale classical simulation of quantum circuits, tensor-network message passing, physics-inspired methods for combinatorial optimization, and exact maximum-likelihood decoding of quantum error-correcting codes. His research has also been highlighted by Science Journal and Phys.org. Beyond academia, Feng maintains active collaborations with leading technology companies, including Google and NVIDIA.