近日，由我院Filippo Fabrocini教授作为通讯作者，Kostas Terzidis教授与两位博士研究生陈得恩和Krid Jinklub作为共同作者，协同其他高校学者团队共同撰写的论文《Quantum Neural Networks and Topological Quantum Field Theories》（《量子神经网络与拓扑量子场论》）在全球重要的神经网络研究期刊《Neural Networks》（《神经网络》）上发表。
该期刊代表了世界上最古老的三个神经网络协会：国际神经网络协会（INNS）、欧洲神经网络协会（ENNS）和日本神经网络协会（JNNS）。2020年《Neural Networks》期刊的影响因子为9.171（5年），该期刊被收录在SCI计算机科学和人工智能、神经科学的Q1中。
在这篇论文中，作者提出了一个已经探索了两年多的新研究范式，为在量子计算平台上实现神经网络开辟了一个全新的方法，将来自拓扑量子场论的概念应用于人工智能。论文旨在表明 (1) 量子神经网络可以被映射到自旋网络上，因此可以依据拓扑量子场论对其进行分析；(2)一些机器学习的关键概念可以通过拓扑量子场论的术语进行重新表述；(3）该框架还通过建立数学或原理性的模型（由一套给定的规则组成），为理解深度神经网络的泛化能力提供了一个可行的假设，尽管这项技术具有出色的表现，但这在当前深度神经网络的研究中仍是缺失的。
自旋网络状态的演化（Evolution of spin network states）菲利波·法布罗基尼（Filippo Fabrocini）博士现任同济大学设计创意学院教授，并担任“可持续人工智能实验室”主任。他还是《Experimental and Theoretical AI》（Taylor & Francis）期刊的副主编。他的研究领域是机器学习、量子机器学习、AI伦理和AI艺术。康思达（Kostas Terzidis）美国密歇根大学博士，现任同济大学设计创意学院教授和"尚想实验室"的主任。他曾是哈佛大学设计研究生院的副教授，撰写了四本关于设计与计算的书籍以及几十篇论文。陈得恩是同济大学设计创意学院的博士生，他感兴趣的领域包括深度学习、进化AI和AI艺术，他的博士论文关于自我复制的神经网络。Krid Jinklub是同济大学设计创意学院的博士生，他正在研究量子神经网络。他在北京理工大学软件学院获得了硕士学位。他的主要兴趣领域包括分布式账本技术和人工智能。
The paper “Quantum Neural Networks and Topological Quantum Field Theories”
co-authored by a team of scholars coming from different universities such as
Fudan University, Yale University, Tufts University, and others includes a
number of D&I scholars as well. Prof. Filippo Fabrocini is the
corresponding author. Prof. Kostas Terzidis is a co-author together with two
PhD students from D&I, Mr. De’en Chen and Mr. Krid Jinklub. The first
author is Prof. Antonino Marciano’ from the Department of Physics at Fudan
University. The paper has been published on the Journal of Neural Networks,
the most important neural networks journal representing the world's three
oldest neural networks societies: the International Neural Network Society(INNS), the European Neural Network Society (ENNS), and the Japanese
Neural Network Society (JNNS). The 2020 Impact Factor of the Journal of
Neural Networks has been extremely high: 9.171 (5 years). The journal is
abstracted and indexed in Scopus and the Science Citation Index.In this article, the authors illustrate a new research paradigm that
has been explored for over two years. In particular, the paper opens a totally
new approach for implementing Neural Networks on Quantum Computing platforms.
The approach makes use of concepts coming from Topological Quantum Field Theory
and moves them to Artificial Intelligence. The paper intends to show that: (1)
Quantum Neural Networks can be mapped onto spin-networks, with the consequence
that the level of analysis of their operation can be carried out on the side of
Topological Quantum Field Theory; (2) A number of Machine Learning key-concepts
can be rephrased by using the terminology of Topological Quantum Field Theories.
(3) Last but not least, the framework also provides a working hypothesis for
understanding the generalization behavior of Deep Neural Networks by adopting a
mathematical or principled model that is still missing behind the undoubtably
amazing achievements of this technology.Prof. Filippo Fabrocini is Professor in the College of Design &
Innovation and Director of the “Sustainable AI Lab”. He is also affiliated to Italy
National Research Center (Institute for Computing Applications) and Deputy
Editor of the “Journal of Experimental and Theoretical AI” (Taylor &
Francis). His main areas of interest are Machine Learning, Quantum Machine
Learning, Ethical AI, and AI Art.Prof. Kostas Terzidis is Professor in the College of Design and
Innovation at Tongji University and the Director of the “Permutation Design Lab”. Previously, he was an Associate Professor at
Harvard University’s Graduate School of Design. He is the author of four
academic books on design and computation and a variety of papers.De'en Chen is currently a PhD student in the College of Design &
Innovation at Tongji University. His research interests include Deep Learning,
Evolutionary Artificial Intelligence, and AI Art. His PhD thesis is about
Self-replicating Neural Networks.Mr. Krid Jinklub is a PhD student in the College of Design &
Innovation at Tongji University where he is working on Quantum Neural Networks.
Previously he got a Master in the School of Software Engineering at Beijing
Institute of Technology. His main areas of interest include Distributed Ledger
Technology and AI.