Introduction
System perception, understanding, prediction and control in specific fields create challenges in real world environments such as the ocean system, urban system, network system and swarm system. Dedicated to achieving a high level of understanding and abstraction of computing architecture in such specific domains, the Domain-Oriented Intelligence System Research Center focuses on integration and improvement of signal processing, network optimization, data processing, model adaptation, intelligent computing and other capabilities.
Based on its strong academic research and technology efforts, the center has already achieved a series of representative achievements in artificial intelligence, data mining, network collaboration and intelligent computing. Its series of technical tools and computing systems mark the center as a visionary organization becoming known for international scientific research and engineering practice ability.
The center is the main supporting department of the national Hua Luogeng Intelligent Data Analysis Center, named in honor of Chinese Academy of Sciences academician Hua Luogeng (1910-1985), internationally famous mathematician and a leader of mathematics research and education in China. An author and the educator of a generation of Chinese mathematicians, he served as director of the preparatory committee of the CAS Institute of Computing Technology. Mathematical accomplishments named for him include Hua’s theorem, Weil-Hua inequation, Hua's inequation, Brauer-Cartan-Hua theorem, Hua matrices and Hua-Wang (Zhonglie) Method.
General Objectives
The center’s mission includes designing the new generation of evolving artificial intelligence that could improve itself in the open world, revealing the mechanism of “signal-information-knowledge-decision,” developing corresponding algorithms, building heterogeneous intelligent computing systems highly adaptable to domain application, creating the tool chain for optimization and deployment of intelligent algorithms and models, and innovating multi-agent communication and collaboration methods.
Research Fields
The Domain-Oriented Intelligent System Research Center has these main research areas:
1.Optimization of Deep Neural Networks
Compression and acceleration of deep neural networks (including knowledge distillation and pruning)
Interpretability of deep neural networks (including those mathematically and semantically interpretable)
Generalization of deep neural networks (including lifelong and reinforcement learning, domain generalization and adaptation)
2.Efficient Deep Learning for Edge Intelligence
Algorithm-hardware co-design for lightweight deep learning
End-to-end AI compiler for heterogeneous AI processors
Edge collaborative resource scheduling methods and systems
3.Embedded AI Computing Systems
Embedded computing architectures and circuits used in multi neural networks accelerating and multi tasks real-time processing
Migration and deployment of deep learning algorithms on embedded computing systems
Hardware/software co-designing for practical embedded deep learning application
4.Remote Sensing Image Processing and Analysis
Synthetic Aperture Radar (SAR) signal processing
Modeling of electromagnetic scattering characteristics of typical objects and identification in SAR image
Feature extraction and object identification in optical remote sensing image
5.Electromagnetic Signal Perception and Jamming Technique
Principles of radar jamming and anti-jamming
Analysis and recognition of radar emitter signals
Deception jamming signal generation against SAR
6.Spatial-Temporal Data（ST Data）Science and Technology
General representation and generation technology for spatial-temporal data
Domain-oriented ST data analysis and prediction, including time series, trajectory, multivariate time series, and spatial-temporal graph
Interdisciplinary research concerning ST data analysis, such as an ST data in transportation system, ST data in environment system, and ST data in maritime system
7.Intelligent Wireless Networks and Swarm Intelligence
Network protocols for hard real-time wireless communication, including MAC, flow scheduling, routing and network coding
Performance evaluation and optimization of wireless networks, including industrial IoT, FANETs, and wireless sensor networks
Swarm intelligence algorithms
Platform design and development for intelligent wireless networks
8.AI-based Network Management
Data fusion and knowledge mining from logs, monitoring data, application data, etc.
Interpretable network anomaly detection and root cause analysis
Automated network control under threat (attack, performance drop, routing failures)
Director：Yongjun Xu
E-mail：xyj at ict.ac.cn
downloadFile