认知学习：电磁频谱空间机器学习新范式
Cognitive Learning: A New Paradigm for Machine Learning in Electromagnetic Spectrum Environment
- 电子学报 2023年51卷第6期 页码：1430-1442
- 作者机构：
南京航空航天大学电磁频谱空间认知动态系统工信部重点实验室，江苏南京 211106
- 作者简介：
[ "阮天宸 男，南京航空航天大学电子信息工程学院/集成电路学院在读博士生.主要研究方向为认知学习、信号源识别与搜索.E-mail: ruantianchen@nuaa.edu.cn" ]
[ "吴启晖（通讯作者） 男，南京航空航天大学特聘教授、副校长，教育部“长江学者”特聘教授，南京航空航天大学电子信息工程学院/集成电路学院教授、博导.主要研究方向为认知科学与应用、电磁空间频谱认知智能管控、无人机认知集群." ]
[ "赵世瑾 女，南京航空航天大学电子信息工程学院/集成电路学院在读博士生.主要研究方向为认知强化学习、视觉导航.E-mail: shijin_zhao@nuaa.edu.cn" ]
[ "周福辉 男，国家优秀青年基金获得者，南京航空航天大学电子信息工程学院/集成电路学院教授、博导.主要研究方向为频谱智能管控和资源鲁棒优化、认知智能与知识图谱.E-mail: zhoufuhui@nuaa.edu.cn" ]
[ "黄洋 男，南京航空航天大学电子信息工程学院/集成电路学院副教授.主要研究方向为强化学习、深度学习、最优化理论、协同智能、频谱管控.E-mail: yang.huang.ceie@nuaa.edu.cn" ]
- 基金信息：国家自然科学基金重点项目(62231015)
DOI：10.12263/DZXB.20221137
中图分类号： TN929.5;TP181收稿：2022-10-10，
修回：2023-03-07，
纸质出版：2023-06-25
- 稿件说明：
移动端阅览
阮天宸,吴启晖,赵世瑾等.认知学习：电磁频谱空间机器学习新范式[J].电子学报,2023,51(06):1430-1442.
RUAN Tian-chen,WU Qi-hui,ZHAO Shi-jin,et al.Cognitive Learning: A New Paradigm for Machine Learning in Electromagnetic Spectrum Environment[J].ACTA ELECTRONICA SINICA,2023,51(06):1430-1442.
阮天宸,吴启晖,赵世瑾等.认知学习：电磁频谱空间机器学习新范式[J].电子学报,2023,51(06):1430-1442. DOI： 10.12263/DZXB.20221137.
RUAN Tian-chen,WU Qi-hui,ZHAO Shi-jin,et al.Cognitive Learning: A New Paradigm for Machine Learning in Electromagnetic Spectrum Environment[J].ACTA ELECTRONICA SINICA,2023,51(06):1430-1442. DOI： 10.12263/DZXB.20221137.
摘要
随着无线设备及其应用呈现指数级增长，以及各种无线技术的出现，电磁频谱环境呈现环境多域、态势多维、应用多样、行为多变、信号密集的复杂特性.通过对海量数据进行即时有效的分析和处理，实现有效的频谱资源管控，对于频谱高效利用和电磁致胜显得尤为重要.以机器学习为代表的智能技术方法为电磁频谱数据分析提供了新的发展机遇，目前其已经应用于无线网络、频谱管理、资源管理等场景.然而，将机器学习应用于电磁频谱空间存在许多挑战.因为电磁频谱空间环境任务动态变化、高质量标记样本稀缺、频谱决策对高实时性的需求、知识利用和迁移困难等难点，现有机器学习算法难以很好地适用于电磁频谱空间的智能认知和决策.以干扰识别的研究为例，传统的技术包括信号特征提取和模式识别两个阶段.随着计算机算力的进步，深度学习技术凭借强大的特征表征能力逐渐占据主导地位.然而电磁频谱空间环境和数据往往存在未知、动态变化等不确定性因素，深度学习难以快速学习和适应未观测到的任务，其泛化性较差，且依赖大量高质量标记样本和人工调整超参数进行离线训练.因此，尽管深度学习在大多数研究场景中展现出了传统机器学习难以望其项背的结果，传统的机器学习图像识别模型仍然在某些场景中具备优势.在干扰识别中，若根据环境和任务特征动态地调度传统机器学习算法和深度学习算法及其超参数，则可以避免单一算法的缺陷，使总体性能得到改善.由此可见，提出一个能够自适应动态环境与任务、不断积累新知识、对错误样本数据具有鲁棒性、部分可解释的学习范式，对于从感知智能向认知智能跨越、迈向通用人工智能具有重要的研究意义和实际应用价值.本文从频谱特性出发，提出了电磁频谱空间机器学习新范式：认知学习，试图在“认知”的层面使用机器学习技术.认知学习范式包括离线训练、在线执行、在线结果反馈、离线自学习过程，对应大脑认知机制的执行控制和内省过程.离线训练探索算法结构和超参数与环境和任务的匹配关系，在线时能够有效地根据实际环境执行快速决策，选择合适的算法和超参数并输出学习结果，同时利用学习结果的反馈传递进行离线自学习的自我优化，形成离线-在线-离线的结构.具体来说，认知特征提取模块建立环境和任务特征与算法类型和超参数之间的匹配关系，从而得到合适的算法类型和超参数；记忆模块调用具体算法和超参数值，进行算法的重构；学习网络模块根据数据执行所选择的算法，输出在线学习结果；认知评估模块对在线学习结果进行内省，调整记忆中的频谱数据库与认知案例库，更新最合适的算法和超参数.在离线阶段，根据当前的学习结果判断是否将其作为新的认知案例存储至记忆空间；认知控制模块重新训练算法和超参数的选择，从而不断优化算法和超参数选择性能.此外，本文揭示了认知学习中的动态匹配定理和最优匹配定理，并根据没有免费午餐定理的思路完成了定理证明.动态匹配定理论证了最优算法和超参数的存在性与动态性，即对于任意环境和任务，总存在一组性能最优的算法和超参数.不同的环境与任务对应不同的最佳算法和超参数，并且本文提出的框架可以为动态环境和任务选择出这组算法和超参数.本文还揭示了最优匹配定理，即当算法和超参数选择的知识增加时，候选的算法类型和超参数搜索空间减小，能够选择到最合适的算法类型和超参数的概率提升.本文设计了一种面向干扰识别的认知学习方案.通过认知特征提取，基于差异性的任务需求和信号模态，对不同问题采取不同算法和超参数产生测试结果，生成认知案例库.将大量认知案例作为训练样本训练算法和超参数选择的神经网络.对于神经网络结构，采用了径向基函数神经网络，其优势是隐藏单元的数量不需要手动设置.当各种问题特征的训练样本数量增加时，隐藏层的大小可以根据近似线性相关准则自动扩展，缓解了神经网络中灾难性遗忘的问题.本文进行了不同数据集和不同识别要求情况下的干扰识别仿真.仿真结果表明，所提框架能适应动态变化的环境和任务、能通过自学习提升性能、能缓解错误标记样本的影响，对电磁频谱空间机器学习的发展具有启发式意义.认知学习的发展仍有进步空间，未来需要从多模态计算、知识的可解释性、大规模任务、学习架构的可扩展性等方向寻求突破.
Abstract
With the exponential growth in the number of wireless devices and wireless applications
and the emergence of various wireless technologies
the electromagnetic spectrum environment has presented complex characteristics such as multi-domain environments and situations
diverse applications
changeable behaviors
and dense signals. Spectrum resource control through immediate and effective analysis and processing of massive data is vital for efficient spectrum utilization and electromagnetic spectrum dominance. Intelligent technology represented by machine learning methods provides new opportunities to analyze electromagnetic spectrum data. Machine learning methods have been applied to wireless networks
spectrum management
resource management
and other scenarios. There are many challenges when applying machine learning to the electromagnetic spectrum environment. Because of the dynamically changing environment and tasks
the scarcity of high-quality labeled samples
real-time spectrum decision-making
and the difficulty of knowledge utilization and transfer in the electromagnetic spectrum environment
existing machine learning algorithms are not ideal for intelligent cognitive decision-making. Taking the research on interference recognition as an example
the traditional methods include two stages: signal feature extraction and pattern recognition. With the growth in computer computing power
deep learning techniques gradually replace traditional methods because of their powerful feature characterization capability. However
the electromagnetic spectrum environment and data often have uncertainties caused by unknown
dynamic changes. Adapting deep learning to unobserved tasks quickly is challenging
indicating poor generalization capability. Furthermore
it is limited by its reliance on a large number of high-quality labeled samples and manual adjustment of hyperparameters for offline training. Thus
although deep learning has far outperformed traditional methods in most research scenarios
traditional machine learning-based image recognition models still have advantages in some scenarios. In interference recognition
if traditional machine learning algorithms and deep learning algorithms and their hyperparameters can be dynamically scheduled according to the environment and task characteristics
the shortcomings of using a single algorithm can be avoided
resulting in improved overall performance. It is necessary to propose a learning paradigm that continuously adapts to dynamic environments and tasks
accumulates new knowledge
is robust to erroneous sample data
and is partially interpretable. The leap from perceptual intelligence to cognitive intelligence toward general artificial intelligence has research significance and practical value.
Given the characteristics of the electromagnetic spectrum
this paper proposes a new paradigm of machine learning method in the electromagnetic spectrum environment—cognitive learning
attempting to use machine learning techniques at a “cognitive” level. It includes offline training
online execution
online feedback and offline self-learning processes
corresponding to the executive control and introspective processes of the brain's cognitive mechanisms. Offline training explores the matching relationship between the algorithm type and hyperparameters with the environment and the task. During the online process
it can perform decision-making based on the actual environment
select appropriate algorithms and hyperparameters
and output learning results. The feedback of learning results can be used for self-optimization in offline self-learning
forming an offline-online-offline structure.The cognitive feature extraction module establishes matching relationships between features of environments and tasks and algorithm types and hyperparameters to obtain the appropriate algorithm types and hyperparameters. The memory module calls specific algorithms and hyperparameter values for algorithm reconstruction. The learning network module executes the selected algorithm based on the input data and outputs online learning results.
The cognitive evaluation module introspects the online learning results and adjusts the data in the memory database and cognitive case base
and updates the most appropriate algorithm and hyperparameters. In the offline stage
the current learning results are judged whether to be stored as a new cognitive case in the memory. The cognitive control module retrains the selection of algorithms and hyperparameters to continuously optimize the performance of selecting algorithms and hyperparameters.
Moreover
we reveal and confirm the dynamic matching theorem and the optimal matching theorem in cognitive learning according to the No Free Lunch Theorem. The dynamic matching theorem states the existence and dynamics of optimal algorithms and hyperparameters. For any environment or task
there is always a set of algorithms and hyperparameters with optimal performance
which differs for different environments and tasks. The proposed cognitive learning paradigm can select appropriate algorithms and hyperparameters. The optimal matching theorem states that as the knowledge of algorithm and hyperparameter selection increases
the search space of candidate algorithm types and hyperparameters decreases
and the probability of selecting the most suitable algorithm type and hyperparameters increases.
A cognitive learning scheme for radio frequency interference identification is designed. Different algorithms and hyperparameters are adopted for different problems to generate test results and a cognitive case base based on the differential task requirements and signal modalities acquired by cognitive feature extraction. A large number of cognitive cases are used as training samples to train the neural network for algorithm and hyperparameter selection. A radial basis function neural network is used for the neural network structure
which has the advantage that the number of hidden units does not need to be set manually. The size of hidden layers can be automatically expanded according to the approximate linear correlation criterion when the number of training samples for various problem features increases
alleviating the problem of catastrophic forgetting in neural networks.
We performed interference recognition simulations with different datasets and different recognition requirements. The simulation results demonstrate that the proposed paradigm can adapt to dynamically changing environments and tasks
improve performance by self-learning
and alleviate the impact of mislabeled samples. The cognitive learning paradigm has a pioneering significance for developing machine learning methods applied to the electromagnetic spectrum environment. There remains room for improvement in cognitive learning. We must seek future research breakthroughs in multimodal computing
knowledge interpretability
large-scale tasks
and learning architecture scalability.
关键词
Keywords
references
吴启晖 , 任敬 . 电磁频谱空间认知新范式: 频谱态势 [J]. 南京航空航天大学学报 , 2016 , 48 ( 5 ): 625 - 632 .
WU Q H , REN J . New paradigm of electromagnetic spectrum space: Spectrum situation [J]. Journal of Nanjing University of Aeronautics & Astronautics , 2016 , 48 ( 5 ): 625 - 632 . (in Chinese)
李京华 , 丁国如 , 徐以涛 , 等 . 面向电磁频谱战的群体智能初探 [J]. 航空兵器 , 2020 , 27 ( 4 ): 56 - 63 .
LI J H , DING G R , XU Y T , et al . Preliminary study on group intelligence for electromagnetic spectrum warfare [J]. Aero Weaponry , 2020 , 27 ( 4 ): 56 - 63 . (in Chinese)
REICHSTEIN M , CAMPS-VALLS G , STEVENS B , et al . Deep learning and process understanding for data-driven Earth system science [J]. Nature , 2019 , 566 ( 7743 ): 195 - 204 .
PEI J , DENG L , SONG S , et al . Towards artificial general intelligence with hybrid Tianjic chip architecture [J]. Nature , 2019 , 572 ( 7767 ): 106 - 111 .
WOLPERT D H , MACREADY W G . No free lunch theorems for optimization [J]. IEEE Transactions on Evolutionary Computation , 1997 , 1 ( 1 ): 67 - 82 .
ELIASMITH C , STEWART T C , CHOO X , et al . A large-scale model of the functioning brain [J]. Science , 2012 , 338 ( 6111 ): 1202 - 1205 .
BEYER L , HÉNAFF O J , KOLESNIKOV A , et al . Are we done with ImageNet? [EB/OL]. ( 2020-06-12 )[ 2022-10-01 ]. https://arxiv.org/abs/2006.07159 https://arxiv.org/abs/2006.07159 .
LIU X , SUN Q Q , LU W D , et al . Big-data-based intelligent spectrum sensing for heterogeneous spectrum communications in 5G [J]. IEEE Wireless Communications , 2020 , 27 ( 5 ): 67 - 73 .
ZENG G X , CHEN Y , CUI B , et al . Continual learning of context-dependent processing in neural networks [J]. Nature Machine Intelligence , 2019 , 1 ( 8 ): 364 - 372 .
TENENBAUM J B , KEMP C , GRIFFITHS T L , et al . How to grow a mind: Statistics, structure, and abstraction [J]. Science , 2011 , 331 ( 6022 ): 1279 - 1285 .
AL-SHAWABKA A , RESTUCCIA F , D'ORO S , et al . Exposing the fingerprint: Dissecting the impact of the wireless channel on radio fingerprinting [C]// IEEE INFOCOM 2020 - IEEE Conference on Computer Communications . Toronto : IEEE , 2020 : 646 - 655 .
MAJUMDER U K , BLASCH E P , GARREN D A . IEEE radar conference 2019 tutorial session I machine learning techniques for radar ATR : Monday , April 22 , 2019 time: 11: 00AM–3: 00PM[C]// 2019 IEEE Radar Conference (RadarConf) . Boston : IEEE , 2019: 1 - 93 .
SUN Y H , PENG M G , ZHOU Y C , et al . Application of machine learning in wireless networks: Key techniques and open issues [J]. IEEE Communications Surveys & Tutorials , 2019 , 21 ( 4 ): 3072 - 3108 .
ZHOU F H , LU G Y , WEN M W , et al . Dynamic spectrum management via machine learning: State of the art, taxonomy, challenges, and open research issues [J]. IEEE Network , 2019 , 33 ( 4 ): 54 - 62 .
HUSSAIN F , HASSAN S ALI , HUSSAIN R , et al . Machine learning for resource management in cellular and IoT networks: Potentials, current solutions, and open challenges [J]. IEEE Communications Surveys & Tutorials , 2020 , 22 ( 2 ): 1251 - 1275 .
方芳 , 李永贵 , 牛英滔 , 等 . 基于决策树算法的干扰信号识别 [J]. 通信技术 , 2019 , 52 ( 11 ): 2617 - 2623 .
FANG F , LI Y G , NIU Y T , et al . Jamming signal recognition based on decision tree algorithm [J]. Communications Technology , 2019 , 52 ( 11 ): 2617 - 2623 . (in Chinese)
徐国进 . 典型通信干扰信号识别技术研究 [D]. 成都 : 电子科技大学 , 2018 .
XU G J . Research on Identification of Typical Communication Jamming Signals [D]. Chengdu : University of Electronic Science and Technology of China , 2018 . (in Chinese)
朱鹏程 . GNSS干扰检测与识别技术研究 [D]. 重庆 : 重庆大学 , 2018 .
ZHU P C . Research on GNSS Interference Detection and Identification Technologies [D]. Chongqing : Chongqing University , 2018 . (in Chinese)
郝万兵 , 马若飞 , 洪伟 . 基于时频特征提取的雷达有源干扰识别 [J]. 火控雷达技术 , 2017 , 46 ( 4 ): 11 - 15 .
HAO W B , MA R F , HONG W . Radar active jamming identification based on time-frequency characteristic extraction [J]. Fire Control Radar Technology , 2017 , 46 ( 4 ): 11 - 15 . (in Chinese)
SHEN A , LIU Y , ZHANG Y , et al . The method of interference recognition in mobile communication network based on deep learning [C]// International Conference on Signal and Information Processing, Networking and Computers . Singapore : Springer , 2019 : 296 - 306 .
练东海 . 基于深度学习的无线通信干扰信号识别与处理 [J]. 通信电源技术 , 2021 , 38 ( 2 ): 104 - 107 .
LIAN D H . Deep learning based wireless communication interference signal recognition and processing [J]. Telecom Power Technology , 2021 , 38 ( 2 ): 104 - 107 . (in Chinese)
党泽 . 基于深度学习的无线通信干扰信号识别与处理技术研究 [D]. 成都 : 电子科技大学 , 2020 .
DANG Z . Research on the Technology of Wireless Communication Interference Signal Identification and Processing Based on Deep Learning [D]. Chengdu : University of Electronic Science and Technology of China , 2020 . (in Chinese)
杨岚 , 肖海涛 , 张渭乐 . 基于频带相关性Deep Learning的无线通信干扰智能识别 [J]. 航空科学技术 , 2022 , 33 ( 4 ): 108 - 118 .
YANG L , XIAO H T , ZHANG W L . Intelligent jamming identification based on deep learning with frequency band correlation for wireless communication [J]. Aeronautical Science & Technology , 2022 , 33 ( 4 ): 108 - 118 . (in Chinese)
刘华祠 . 基于传统机器学习与深度学习的图像分类算法对比分析 [J]. 电脑与信息技术 , 2019 , 27 ( 5 ): 12 - 15 .
LIU H C . Comparative analysis of image classification algorithms based on traditional machine learning and deep learning [J]. Computer and Information Technology , 2019 , 27 ( 5 ): 12 - 15 . (in Chinese)
KERSCHKE P , HOOS H H , NEUMANN F , et al . Automated algorithm selection: Survey and perspectives [J]. Evolutionary Computation , 2019 , 27 ( 1 ): 3 - 45 .
崔建双 , 刘晓婵 , 杨美华 , 等 . 基于元学习推荐的优化算法自动选择框架与实证分析 [J]. 计算机应用 , 2017 , 37 ( 4 ): 1105 - 1110 .
CUI J S , LIU X C , YANG M H , et al . Meta-learning based optimization algorithm selection framework and its empirical study [J]. Journal of Computer Applications , 2017 , 37 ( 4 ): 1105 - 1110 . (in Chinese)
KHAN I , ZHANG X C , REHMAN M , et al . A literature survey and empirical study of meta-learning for classifier selection [J]. IEEE Access , 2020 , 8 : 10262 - 10281 .
任义 , 迟翠容 , 单菁 , 等 . 基于元学习的推荐算法选择优化框架实证 [J]. 计算机工程与设计 , 2020 , 41 ( 6 ): 1610 - 1616 .
REN Y , CHI C R , SHAN J , et al . Empirical study on recommendation algorithm selection optimization framework based on meta-learning [J]. Computer Engineering and Design , 2020 , 41 ( 6 ): 1610 - 1616 . (in Chinese)
WU Q H , RUAN T C , ZHOU F H , et al . A unified cognitive learning framework for adapting to dynamic environments and tasks [J]. IEEE Wireless Communications , 2021 , 28 ( 6 ): 208 - 216 .
BRITTO A S , SABOURIN R , OLIVEIRA L E S . Dynamic selection of classifiers—A comprehensive review [J]. Pattern Recognition , 2014 , 47 ( 11 ): 3665 - 3680 .
BOHN B , GRIEBEL M , RIEGER C . A representer theorem for deep kernel learning [EB/OL]. ( 2017-09-29 )[ 2022-10-01 ]. https://arxiv.org/abs/1709.10441 https://arxiv.org/abs/1709.10441 .
KAELBLING L P . The foundation of efficient robot learning [J]. Science , 2020 , 369 ( 6506 ): 915 - 916 .
0
浏览量
27
下载量
2
CSCD
- 网络PDF下载
- Meta-XML下载
- BibTeX下载
- Endnote下载
- RefMan 下载
- NoteExpress下载
- NoteFirst下载
关联资源
相关文章
相关作者
相关机构
- 技术支持由北京北大方正电子有限公司提供 京ICP备09064830号-19
京公网安备11010802024621 - 本系统建议在Chrome、 IE9+ 以上版本浏览器阅读本站内容，360浏览器请切换至极速模式
- Cookies帮助我们提供服务并提供个性化体验。使用本网站，即表示您同意我们使用Cookies