Liquid Neural Networks: Next-Generation AI for Telecom from First Principles

doi:10.12142/ZTECOM.202502008

ZTE Communications ›› 2025, Vol. 23 ›› Issue (2): 76-84.DOI: 10.12142/ZTECOM.202502008

• Review • Previous Articles Next Articles

Liquid Neural Networks: Next-Generation AI for Telecom from First Principles

ZHU Fenghao, WANG Xinquan, ZHU Chen, HUANG Chongwen()

Zhejiang University, Hangzhou 310027, China

Received:2025-04-01 Online:2025-06-25 Published:2025-06-10
About author:ZHU Fenghao received his BE degree in information engineering from Zhejiang University, China in 2023, and he is currently pursuing his MS degree with the College of Information Science and Electronic Engineering, Zhejiang University. His current research interests include massive MIMO, signal processing, and machine learning. He is a recipient of 2024 IEEE ComSoc Conference Travel Grant.
WANG Xinquan is currently pursuing his BE degree at Zhejiang University, China. His current research interests include 6G, beamforming and machine learning. He is a recipient of 2024 IEEE ComSoc Student Travel Grant.
ZHU Chen received his BS degree from North University of China in 2010, and MS degree from Zhejiang University of Technology, China in 2013. He is currently engaged in teaching and researching at the College of Engineering, Zhejiang University, China. His main research interests include general sense computing integration, machine learning, image processing, and cloud-edge collaborative computing.
HUANG Chongwen (chongwenhuang@zju.edu.cn) obtained his BS degree in 2010 from Nankai University, China and MS degree from the University of Electronic Science and Technology of China in 2013, and PhD degree from Singapore University of Technology and Design (SUTD) in 2019. From Oct. 2019 to Sep. 2020, he is a Postdoc at SUTD. Since Sep. 2020, he joined Zhejiang University as a tenure-track young professor. Prof. HUANG is the recipient of 2021 IEEE Marconi Prize Paper Award, 2023 IEEE Fred W. Ellersick Prize Paper Award and 2021 IEEE ComSoc Asia-Pacific Outstanding Young Researcher Award. He has served as an editor of IEEE Communications Letter, Elsevier Signal Processing, EURASIP Journal on Wireless Communications and Networking and Physical Communication since 2021. His main research interests focus on holographic MIMO surface/reconfigurable intelligent surface, B5G/6G wireless communications, mmWave/THz Communications, deep learning technologies for Wireless communications, etc.
Supported by:
the China National Key R&D Program(2021YFA1000500);National Natural Science Foundation of China(62331023);Zhejiang Provincial Natural Science Foundation of China(LR22F010002);Zhejiang Provincial Science and Technology Plan Project(2024C01033)

Abstract

Abstract:

Recently, a novel type of neural networks, known as liquid neural networks (LNNs), has been designed from first principles to address robustness and interpretability challenges facing artificial intelligence (AI) solutions. The potential of LNNs in telecommunications is explored in this paper. First, we illustrate the mechanisms of LNNs and highlight their unique advantages over traditional networks. Then we explore the opportunities that LNNs bring to future wireless networks. Furthermore, we discuss the challenges and design directions for the implementation of LNNs. Finally, we summarize the performance of LNNs in two case studies.

Key words: artificial intelligence (AI), liquid neural networks (LNNs), telecommunications, wireless networks

ZHU Fenghao, WANG Xinquan, ZHU Chen, HUANG Chongwen. Liquid Neural Networks: Next-Generation AI for Telecom from First Principles[J]. ZTE Communications, 2025, 23(2): 76-84.

Figures/Tables 6

Figure 1 6G communication scenarios with AI integration

Figure 2 Liquid neuron and the ODE modeling

Figure 3 Structure of a CfC neuron and a four-layer NCP

Table 1 Simulation parameters in Figs. 4 and 5

Parameters	Fig. 4	Fig. 5
BS Antenna number	4	64
BS Antenna spacing	0.5 $λ$	0.5 $λ$
User number	1	4
User antenna number	1	2
Central frequency	6 GHz	28 GHz
Liquid neuron number	/	30

Table 1 Simulation parameters in Figs. 4 and 5

Parameters	Fig. 4	Fig. 5
BS Antenna number	4	64
BS Antenna spacing	0.5 $λ$	0.5 $λ$
User number	1	4
User antenna number	1	2
Central frequency	6 GHz	28 GHz
Liquid neuron number	/	30

Figure 4 MSE versus prediction length with real-world channel state information

Figure 5 Average SE in a dynamic beamforming scenario

References 25

1	WANG C X, YOU X H, GAO X Q, et al. On the road to 6G: visions, requirements, key technologies, and testbeds [J]. IEEE communications surveys and tutorials, 2023, 25(2): 905–974. DOI: 10.1109/COMST.2023.3249835
2	SHAHID A, KLIKS A, AL-TAHMEESSCHI A, et al. Large-scale AI in telecom: charting the roadmap for innovation, scalability, and enhanced digital experiences [EB/OL]. [2025-02-16].
3	O’SHEA T, HOYDIS J. An introduction to deep learning for the physical layer [J]. IEEE transactions on cognitive communications and networking, 2017, 3(4): 563–575. DOI: 10.1109/TCCN.2017.2758370
4	ZHU F H, WANG X Q, HUANG C W, et al. Beamforming inferring by conditional WGAN-GP for holographic antenna arrays [J]. IEEE wireless communications letters, 2024, 13(7): 2023–2027. DOI: 10.1109/LWC.2024.3402102
5	DAI L L, JIAO R C, ADACHI F, et al. Deep learning for wireless communications: an emerging interdisciplinary paradigm [J]. IEEE wireless communications, 2020, 27(4): 133–139. DOI: 10.1109/MWC.001.1900491
6	ZHU F H, WANG X Q, HUANG C W, et al. Robust beamforming for RIS-aided communications: gradient-based manifold meta learning [J]. IEEE transactions on wireless communications, 2024, 23(11): 15945–15956. DOI: 10.1109/TWC.2024.3435023
7	WANG T H, XIAO W, SEYDE T, et al. Measuring interpretability of neural policies of robots with disentangled representation [EB/OL]. (2023-12-11) [2025-02-16].
8	JIANG W, HAN B, HABIBI A M, et al. The road towards 6G: a comprehensive survey [J]. IEEE open journal of the Communications Society, 2021, 2: 334–366. DOI: 10.1109/OJCOMS.2021.3057679
9	ZHU F H, WANG B H, YANG Z H, et al. Robust millimeter beamforming via self-supervised hybrid deep learning [C]//The 31st European Signal Processing Conference (EUSIPCO). IEEE, 2023: 915–919. DOI: 10.23919/EUSIPCO58844.2023.10289989
10	HASANI R, LECHNER M, AMINI A, et al. Liquid time-constant networks [J]. Proceedings of the AAAI conference on artificial intelligence, 2021, 35(9): 7657–7666. DOI: 10.1609/aaai.v35i9.16936
11	LECHNER M, HASANI R, AMINI A, et al. Neural circuit policies enabling auditable autonomy [J]. Nature machine intelligence, 2020, 2(10): 642–652. DOI: 10.1038/s42256-020-00237-3
12	HASANI R, LECHNER M, AMINI A. Closed-form continuous-time neural networks [J]. Nature machine intelligence, 2022, 4(11): 992–1003. DOI: 10.1038/s42256-022-00556-7
13	ELDAN R, SHAMIR O. The power of depth for feedforward neural networks [C]//The 29th Annual Conference on Learning Theory. PMLR, 2016: 907–940
14	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks [C]//The 26th International Conference on Neural Information Processing Systems. ACM, 2012: 1097–1105. DOI: 10.1145/3065386
15	YU Y, SI X S, HU C H, et al. A review of recurrent neural networks: LSTM cells and network architectures [J]. Neural computation, 2019, 31(7): 1235–1270. DOI: 10.1162/neco_a_01199
16	MA K, ZHANG F, TIAN W Q, et al. Continuous-time mmWave beam prediction with ODE-LSTM learning architecture [J]. IEEE wireless communications letters, 2023, 12(1): 187–191. DOI: 10.1109/LWC.2022.3221159
17	CHEN R T, RUBANOVA Y, BETTENCOURT J, et al. Neural ordinary differential equations [C]//The 32nd International Conference on Neural Information Processing Systems. ACM, 2018: 6572–6583
18	CHAHINE M, HASANI R, KAO P, et al. Robust flight navigation out of distribution with liquid neural networks [J]. Science robotics, 2023, 8(77): eadc8892. DOI: 10.1126/scirobotics.adc8892
19	WANG X Q, ZHU F H, ZHOU Q Y, et al. Energy-efficient beamforming for RISs-aided communications: gradient based meta learning [C]//International Conference on Communications. IEEE, 2024: 3464–3469. DOI: 10.1109/ICC51166.2024.10622978
20	ZHU F H, WANG X Q, HUANG C W, et al. Robust continuous-time beam tracking with liquid neural network [C]//IEEE Global Communications Conference. IEEE, 2024: 4878–4883. DOI: 10.1109/GLOBECOM52923.2024.10900942
21	WEI Z Q, ZHANG Y J, JI D N, et al. Sensing and communication integrated fast neighbor discovery for UAV networks [J]. ZTE Communications, 2024, 22(3): 69–82. DOI: 10.12142/ZTECOM.202403009
22	DU R L, WEI Z Q, YANG Z. Integrated sensing and communication: who benefits more? [J]. ZTE Communications, 2024, 22(3): 37–47. DOI: 10.12142/ZTECOM.202403006
23	WANG B H, ZHU F H, LIU M B, et al. Multi-sources information fusion learning for multi-points NLOS localization [C]//The 99th Vehicular Technology Conference (VTC2024-Spring). IEEE, 2024: 1–6. DOI: 10.1109/VTC2024-Spring62846.2024.10683036
24	YIN H, ZHOU Y H, CAO L, et al. Channel prediction with liquid time-constant networks: an online and adaptive approach [C]//The 94th Vehicular Technology Conference (VTC2021-Fall). IEEE, 2021: 1–6. DOI: 10.1109/vtc2021-fall52928.2021.9625323
25	WANG X Q, ZHU F H, HUANG C W, et al. Robust beamforming with gradient-based liquid neural network [J]. IEEE wireless communications letters, 2024, 13(11): 3020–3024. DOI: 10.1109/lwc.2024.3436576

Liquid Neural Networks: Next-Generation AI for Telecom from First Principles

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 6

References 25

Related Articles 1

Recommended Articles 0

Metrics