• TabularNet: A Neural Network Architecture for Understanding Semantic Structures of Tabular Data.

    Lun Du, Fei Gao, Xu Chen, Ran Jia, Junshan Wang,et al; KDD (2021)

    To simultaneously extract spatial and relational information from tables, we propose a novel neural network architecture, TabularNet. The spatial encoder of TabularNet utilizes the row/column-level Pooling and the Bidirectional Gated Recurrent Unit (Bi-GRU) to capture statistical information and local positional correlation, respectively. For relational information, we design a new graph construction method based on the WordNet tree and adopt a Graph Convolutional Network (GCN) based encoder that focuses on the hierarchical and paratactic relationships between cells. Our neural network architecture can be a unified neural backbone for different understanding tasks and utilized in a multitask scenario. We conduct extensive experiments on three classification tasks with two real-world spreadsheet data sets, and the results demonstrate the effectiveness of our proposed TabularNet over state-of-the-art baselines.

  • Automated Discovery of Interactions and Dynamics for Large Networked Dynamical Systems

    Yan Zhang, Yu Guo, Zhang Zhang, Mengyuan Chen, Shuo Wang, Jiang Zhang, 2021

    Based on a Bernoulli network generator and a Markov dynamics learner, this paper proposes a unified framework for Automated Interaction network and Dynamics Discovery (AIDD) on various network structures and different types of dynamics. The experiments show that AIDD can be applied on large systems with thousands of nodes. AIDD can not only infer the unknown network structure and states for hidden nodes but also can reconstruct the real gene regulatory network based on the noisy, incomplete, and being disturbed data which is closed to real situations. We further propose a new method to test data-driven models by experiments of control. We optimize a controller on the learned model, and then apply it on both the learned and the ground truth models. The results show that both of them behave similarly under the same control law, which means AIDD models have learned the real network dynamics correctly.

  • scCapsNet: a deep learning classifier with the capability of interpretable feature extraction, applicable for single cell RNA data analysis

    Lifei Wang,Rui Nie,Zeyang Yu, Ruyue Xin, Caihong Zheng, Zhang Zhang, Jiang Zhang, Jun Cai; Nature Machine Intelligence, 2: 693-703(2020)

    The scCapsNet model retains the capsule parts of CapsNet but replaces the part of convolutional neural networks with several parallel fully connected neural networks. We apply scCapsNet to scRNA-seq data. The results show that scCapsNet performs well as a classifier and also that the parallel fully connected neural networks function like feature extractors as we supposed.

  • A General Deep Learning Framework for Network Reconstruction and Dynamics Learning

    Zhang Zhang,Yi Zhao,Jing Liu, Shuo Wang, Ruyi Tao, Ruyue Xin, Jiang Zhang; Applied Network Science, 4, 110 (2019)

    A New paper is published on Nature Machine Intelligence. We propose an interpretable deep-learning architecture using capsule networks (called scCapsNet). By utilizing competitive single-cell-type recognition, the scCapsNet model is able to perform feature selection to identify groups of genes encoding different subcellular types. The RNA expression signatures, which enable subcellular-type recognition, are effectively integrated into the parameter matrices of scCapsNet. This characteristic enables the discovery of gene regulatory modules in which genes interact with each other and are closely related in function, but present distinct expression patterns.

  • Simple spatial scaling rules behind complex cities

    Ruiqi Li,Lei Dong,Jiang Zhang, Xinran Wang, Wen-xu Wang, E.G. Stanley;Nature Communications 8, 1841 (2017)

    By a simple model mainly based on spatial attraction and matching growth mechanisms, we reveal that the spatial scaling rules of these three elements are in a consistent framework, which allows us to use any single observation to infer the others. All numerical and theoretical results are consistent with empirical data from ten representative cities. In addition, our model can also provide a general explanation of the origins of the universal super- and sub-linear aggregate scaling laws and accurately predict kilometre-level socioeconomic activity. Our work opens a new avenue for uncovering the evolution of cities in terms of the interplay among urban elements, and it has a broad range of applications

Machine Learning on Complex Systems

Machine learning provides us new tools to handle complexity in big data and solve hard problems. For example, we have applied CapsuleNet model on single-cell RNA-sequencing data to identify cell types. We also applied convolutional network on complex network classifications.

Automated Modelling Complex Systems

Building models for complex systems depends on professional experience which is very hard to be generalized. How can we build models for complex systems according to the behaviors of the system in an automatic way? As new techniques like Graph Network, Deep Learning, and other techniques, we can reconstruct the dynamical models from time series data. We have developed our own models on data-driven modelling of complex systems. Our models can not only learn the dynamics and make predictions, but also can reaveal the hidden network structures hehind the system.

Open Flow Networks

Flows are ubiquitous in open complex systems. For example, energy flow, money flow, material flow, etc. However, there is no appropriate tool to analyze these flows on open systems. We develop one of the network analysis from ecology, which is called open flow networks. By modeling a flow system as a directed weighted network, we can depict distributions and structures of flows in a system. By adding two special nodes, source and sink, into the network, we can model open systems. We have applied this method on different fields including, online education and learning, online forum, web sites, international trading system, food webs, etc.

Scaling Analysis

Many complex systems exhibit universal scaling laws, for example, the famous Kleiber law in biology, and the scaling law in cities. On one side, by discovering scaling laws on the aggregated level from the data, scaling analysis can characterize the macroscopic universal patterns of complex systems, on the other side, scaling laws can provide some insights for the micro-level mechanisms. I have done several works on scaling laws, for example, in urban system, firms, web forums, and so on.


  • 集智俱乐部:走近2050——注意力、互联网与人工智能,人民邮电出版社,2016.6

    随着人工智能程序AlphaGo以4∶1的大比分战胜人类围棋世界冠军李世石,机器将征服人 类的担忧正在甚嚣尘上。《走近 2050》则为我们描绘了一幅人机和谐共生、协同演化的全新场景。 在看得见的未来,人类将越来越多地沉浸于五花八门的虚拟世界以获取各式体验,与此同时,我 们将心甘情愿地将自己的注意力源源不断地输入给机器世界以促使它们进化。
    本书从注意力的角度解读了包括互联网、人工智能、注意力经济、众包、人类计算、计算机 游戏、虚拟现实在内的技术领域及其对社会生活的影响,还创造性地提出了一系列全新的概念: 占意理论、“游戏+”时代、意本家、自动游戏设计、自动化创业、占意通货、许愿树,等等。 所有这些将为我们理解技术与人类的关系、透视人类社会的未来发展和走向提供深刻的洞察。 本书适用于互联网及人工智能从业人员,企业高管,以及对人类与科技的未来、科技如何影 响社会等问题感兴趣的读者。

  • 集智俱乐部:科学的极致——漫谈人工智能,人民邮电出版社,2008.8

    最受欢迎的科学探索群体——集智俱乐部首部科普著作全面涵盖人机交互、脑科学、计算心理学、系统科学、社会科学等前沿知识生动的文笔,精美的插图,数十位科学爱好者带你展开一场人工智能探索之旅。 制造出能够像人类一样思考的机器是人们长期以来的伟大梦想,也是当今科学发展的极致。从《终结者》《黑客帝国》《机器人瓦利》再到《超验骇客》,我们多数人对人工智能的认识还停留在好莱坞电影阶段,然而,人工智能作为一门计算机科学的分支究竟是什么样的?目前发展到了什么阶段?能够战胜人类的终极AI机器真的会存在吗?


Research Group


School of Systems Science
Beijing Normal University, 100875
Beijing, China