Zum Hauptinhalt springen
Artikel
Lizenziert
Nicht lizenziert Erfordert eine Authentifizierung

Model compression for neural network controllers: a tutorial survey with a focus on controllers with latent state space

  • Ganesh Sundaram received a Bachelor’s degree in mechanical engineering from the University of Kerala, India. He obtained an M.Sc. in commercial vehicle technology from the Graduate School of Commercial Vehicle Technology and an M.Sc. in automation and control from the Department of Electrical and Computer Engineering at the University of Kaiserslautern, Germany, in 2020. Since 2020, he has been a Researcher at the Institute of Electromobility, Department of Electrical and Computer Engineering, RPTU University of Kaiserslautern-Landau, Germany. His research interests encompass neural network controllers, embedded hardware, and model compression.

     EMAIL logo     ,

    Jonas Ulmen received his Bachelor’s and Master’s degrees in industrial engineering and management from the School of Business and Economics, University of Kaiserslautern, Germany, in 2017 and 2020. Since 2020, he has been a Researcher at the Institute of Electromobility, Department of Electrical and Computer Engineering, at RPTU University of Kaiserslautern-Landau, Germany. His research focuses on reinforcement learning, world models, self-supervised learning, and model predictive control.

         und

    Daniel Görges received the Dipl.-Ing. and Dr.-Ing. degrees from the Department of Electrical and Computer Engineering, University of Kaiserslautern, Kaiserslautern, Germany, in 2005 and 2011, respectively. He has held positions as a Researcher and Junior Professor at the University of Kaiserslautern and as a Research Group Leader at the German Research Center for Artificial Intelligence (DFKI). Since 2021, he has been a Professor and Head of the Institute of Electromobility in the Department of Electrical and Computer Engineering at RPTU University of Kaiserslautern-Landau. He is chair of the VDI/VDE-GMA Technical Committee 2.15, ”Fundamentals of Networked and Learning Systems,” and co-chair of the VDI/VDE-GMA Technical Area 2, ”Methods of Automation”. His research interests include learning methodologies and distributed control, with applications in electric, automated, and connected driving, as well as robotics, mechatronics, and power systems.
Veröffentlicht/Copyright: 9. März 2026
at - Automatisierungstechnik
Aus der Zeitschrift at - Automatisierungstechnik Band 74 Heft 3

Abstract

Model compression is key for deploying Deep Neural Networks on resource-constrained hardware. Its application to Neural Network Controllers (NNCs) is challenging because it can compromise control-theoretic properties and performance, a critical issue for modern controllers that use latent-space models. This paper surveys and empirically evaluates compression techniques for these models, using an MNIST autoencoder and a Temporal Difference Model Predictive Control agent as test cases across diverse hardware. We find that general compression techniques apply to latent-space models and that careful compression can preserve the theoretical properties of NNCs. Specific findings indicate that quantization can increase latency on non-specialized hardware, fine-tuning is crucial for performance recovery, and hybrid methods yield the best trade-offs.

Zusammenfassung

Die Modellkompression ist entscheidend für den Einsatz tiefer neuronaler Netze auf ressourcenbeschränkter Hardware. Ihre Anwendung auf neuronale Regelungskonzepte (Neural Network Controllers, NNCs) ist jedoch anspruchsvoll, da sie die Regelgüte beeinträchtigen kann – eine besondere Herausforderung für moderne Regler, die latente Zustandsraummodelle verwenden. Dieser Beitrag evaluiert Kompressionsverfahren für solche Modelle anhand zweier Testfälle auf unterschiedlichen Hardwareplattformen: einem MNIST-Autoencoder und einem auf Temporal Difference Model Predictive Control basierenden Agenten. Die Ergebnisse zeigen, dass allgemeine Kompressionsverfahren effektiv auf latente Zustandsraummodelle anwendbar sind und dass eine sorgfältige Kompression die Leistungsfähigkeit der NNCs erhalten kann. Konkret verdeutlichen die Experimente, dass Quantisierung auf nicht spezialisierter Hardware die Latenz erhöhen kann, dass Feinabstimmung (Fine-Tuning) für die Wiederherstellung der Performanz essenziell ist und dass hybride Verfahren die besten Kompromisse bieten.

Keywords: model compression; deep neural networks; latent state space; quantization; pruning; knowledge distillation
Schlagwörter: Modellkompression; Tiefe Neuronale Netzwerke; Latenter Zustandsraum; Quantisierung; Pruning; Wissensdestillation
Corresponding author: Ganesh Sundaram, Institute of Electromobility, RPTU University Kaiserslautern-Landau, 67663 Kaiserslautern, Germany, E-mail: 

About the authors

Ganesh Sundaram

Ganesh Sundaram received a Bachelor’s degree in mechanical engineering from the University of Kerala, India. He obtained an M.Sc. in commercial vehicle technology from the Graduate School of Commercial Vehicle Technology and an M.Sc. in automation and control from the Department of Electrical and Computer Engineering at the University of Kaiserslautern, Germany, in 2020. Since 2020, he has been a Researcher at the Institute of Electromobility, Department of Electrical and Computer Engineering, RPTU University of Kaiserslautern-Landau, Germany. His research interests encompass neural network controllers, embedded hardware, and model compression.

Jonas Ulmen

Jonas Ulmen received his Bachelor’s and Master’s degrees in industrial engineering and management from the School of Business and Economics, University of Kaiserslautern, Germany, in 2017 and 2020. Since 2020, he has been a Researcher at the Institute of Electromobility, Department of Electrical and Computer Engineering, at RPTU University of Kaiserslautern-Landau, Germany. His research focuses on reinforcement learning, world models, self-supervised learning, and model predictive control.

Daniel Görges

Daniel Görges received the Dipl.-Ing. and Dr.-Ing. degrees from the Department of Electrical and Computer Engineering, University of Kaiserslautern, Kaiserslautern, Germany, in 2005 and 2011, respectively. He has held positions as a Researcher and Junior Professor at the University of Kaiserslautern and as a Research Group Leader at the German Research Center for Artificial Intelligence (DFKI). Since 2021, he has been a Professor and Head of the Institute of Electromobility in the Department of Electrical and Computer Engineering at RPTU University of Kaiserslautern-Landau. He is chair of the VDI/VDE-GMA Technical Committee 2.15, ”Fundamentals of Networked and Learning Systems,” and co-chair of the VDI/VDE-GMA Technical Area 2, ”Methods of Automation”. His research interests include learning methodologies and distributed control, with applications in electric, automated, and connected driving, as well as robotics, mechatronics, and power systems.

  1. Research ethics: Not applicable.

  2. Informed consent: Not applicable.

  3. Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

  4. Use of Large Language Models, AI and Machine Learning Tools: Grammarly, including the Writefull extension, was used solely for grammar checking and language polishing. No AI tools were employed to generate or alter original content.

  5. Conflict of interest: Authors state no conflict of interest.

  6. Research funding: None.

  7. Data availability: Not applicable – this is a survey paper with analysis based on publicly available benchmark models.

References

[1] Y. Cheng, D. Wang, P. Zhou, and T. Zhang, “A survey of model compression and acceleration for deep neural networks,” arXiv:1710.09282, 2017.Suche in Google Scholar

[2] S. Han, J. Pool, J. Tran, and W. Dally, “Learning both weights and connections for efficient neural network,” Adv. Neural Inf. Process. Syst., vol. 28, pp. 1135–1143, 2015.Suche in Google Scholar

[3] W. Wen, C. Wu, Y. Wang, Y. Chen, and H. Li, “Learning structured sparsity in deep neural networks,” Adv. Neural Inf. Process. Syst., vol. 29, pp. 2074–2082, 2016.Suche in Google Scholar

[4] Z. Liu, J. Li, Z. Shen, G. Huang, S. Yan, and C. Zhang, “Learning efficient convolutional networks through network slimming,” in Proceedings of the 15th IEEE International Conference on Computer Vision, Venice, Italy (Palazzo del Cinema), IEEE Computer Society, 2017, pp. 2736–2744.10.1109/ICCV.2017.298Suche in Google Scholar

[5] Y. He, X. Zhang, and J. Sun, “Channel pruning for accelerating very deep neural networks,” in Proceedings of the 15th IEEE International Conference on Computer Vision, Venice, Italy (Palazzo del Cinema), IEEE Computer Society, 2017, pp. 1389–1397.10.1109/ICCV.2017.155Suche in Google Scholar

[6] D. Molchanov, A. Ashukha, and D. Vetrov, “Variational dropout sparsifies deep neural networks,” in Proceedings of the 34th International Conference on Machine Learning, Sydney, Australia, Proceedings of Machine Learning Research (PMLR) / JMLR, 2017, pp. 2498–2507.Suche in Google Scholar

[7] J. Frankle and M. Carbin, “The lottery ticket hypothesis: Finding sparse, trainable neural networks,” in Proceeding of the 7th International Conference on Learning Representations, Louisiana, USA, OpenReview.net and Curran Associates, Inc., 2019.Suche in Google Scholar

[8] D. Mocanu, E. Mocanu, P. Stone, P. Nguyen, M. Gibescu, and A. Liotta, “Scalable training of artificial neural networks with adaptive sparse connectivity inspired by network science,” Nat. Commun., vol. 9, 2018, https://doi.org/10.1038/s41467-018-04316-3.Suche in Google Scholar PubMed PubMed Central

[9] G. Sundaram, J. Ulmen, and D. Görges, “Enhanced pruning strategy for multi-component neural architectures using component-aware graph analysis,” in Proceedings of the 1st IFAC Joint Conference on Computers, Cognition and Communication, Padova, Italy, Elsevier, 2025.10.1016/j.ifacol.2025.11.934Suche in Google Scholar

[10] B. Jacob et al.., “Quantization and training of neural networks for efficient integer-arithmetic-only inference,” in Proceedings of the 31st IEEE Conference on Computer Vision and Pattern Recognition, Utah, USA, IEEE Computer Society, 2018, pp. 2704–2713.10.1109/CVPR.2018.00286Suche in Google Scholar

[11] R. Krishnamoorthi, “Quantizing deep convolutional networks for efficient inference: A whitepaper,” arXiv:1806.08342, 2018.Suche in Google Scholar

[12] M. Courbariaux, I. Hubara, D. Soudry, R. El-Yaniv, and Y. Bengio, “Binarized neural networks: Training deep neural networks with weights and activations constrained to+ 1 or-1,” arXiv:1602.02830, 2016.Suche in Google Scholar

[13] K. Wang, Z. Liu, Y. Lin, J. Lin, and S. Han, “AMC: AutoML for model compression and acceleration on mobile devices,” in Proceedings of the 15th European Conference on Computer Vision, Munich, Germany, Springer International, 2019, pp. 784–800.Suche in Google Scholar

[14] S. K. Esser, J. L. McKinstry, D. Bablani, R. Appuswamy, and D. S. Modha, “Learned step size quantization,” in Proceedings of the 8th International Conference on Learning Representations, Addis Ababa, Ethiopia, OpenReview.net, 2020.Suche in Google Scholar

[15] M. Rastegari, V. Ordonez, J. Redmon, and A. Farhadi, “XNOR-Net: Imagenet classification using binary convolutional neural networks,” in Proceedings of the 14th European Conference on Computer Vision, Amsterdam, The Netherlands, Springer International, 2016, pp. 525–542.10.1007/978-3-319-46493-0_32Suche in Google Scholar

[16] G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv:1503.02531, 2015.Suche in Google Scholar

[17] A. Romero, N. Ballas, S. E. Kahou, A. Chassang, C. Gatta, and Y. Bengio, “FitNets: Hints for thin deep nets,” in Proceedings of the 3rd International Conference on Learning Representations, San Diego, California, USA, Computational and Biological Learning Society (CBLS), 2015.Suche in Google Scholar

[18] S. Zagoruyko and N. Komodakis, “Paying more attention to attention: Improving the performance of convolutional neural networks via attention transfer,” in Proceedings of the 15th International Conference on Learning Representations, Toulon, France, OpenReview.net, 2017.Suche in Google Scholar

[19] Y. Zhang, T. Xiang, T. M. Hospedales, and H. Lu, “Deep mutual learning,” in Proceedings of the 31st IEEE Conference on Computer Vision and Pattern Recognition, Utah, USA, The IEEE Computer Society and the Computer Vision Foundation, 2018, pp. 4320–4328.10.1109/CVPR.2018.00454Suche in Google Scholar

[20] T. Furlanello, Z. Lipton, M. Tschannen, L. Itti, and A. Anandkumar, “Born again neural networks,” in Proceedings of the 35th International Conference on Machine Learning, Stockholm, Sweden, Proceedings of Machine Learning Research (PMLR), 2018, pp. 1607–1616.Suche in Google Scholar

[21] D. Chen, J.-P. Mei, C. Wang, Y. Feng, and C. Chen, “Improved knowledge distillation via teacher assistant,” in Proceedings of the 34th AAAI Conference on Artificial Intelligence, vol. 34, New York, USA, Association for the Advancement of Artificial Intelligence (AAAI), 2020, pp. 5191–5198.10.1609/aaai.v34i04.5963Suche in Google Scholar

[22] E. L. Denton, W. Zaremba, J. Bruna, Y. LeCun, and R. Fergus, “Exploiting linear structure within convolutional networks for efficient evaluation,” Adv. Neural Inf. Process. Syst., vol. 27, pp. 1269–1277, 2014.Suche in Google Scholar

[23] M. Jaderberg, A. Vedaldi, and A. Zisserman, “Speeding up convolutional neural networks with low rank expansions,” in British Machine Vision Conference, Nottingham, UK, BMVA Press, 2014.10.5244/C.28.88Suche in Google Scholar

[24] V. Lebedev, Y. Ganin, M. Rakhuba, I. Oseledets, and V. Lempitsky, “Speeding-up convolutional neural networks using fine-tuned CP-decomposition,” in Proceedings of the 3rd International Conference on Learning Representations, San Diego, California, USA, International Conference on Learning Representations (ICLR), 2015.Suche in Google Scholar

[25] Y.-D. Kim, E. Park, S. Yoo, T. Choi, L. Yang, and D. Shin, “Compression of deep convolutional neural networks for fast and low power mobile applications,” in Proceedings of the 15th International Conference on Learning Representations, San Juan, Puerto Rico, OpenReview, 2016.Suche in Google Scholar

[26] X. Yu, T. Liu, X. Wang, and D. Tao, “On compressing deep models by low rank and sparse decomposition,” in Proceedings of the 30th IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, Hawaii, IEEE Computer Society, 2017, pp. 7370–7379.10.1109/CVPR.2017.15Suche in Google Scholar

[27] A. C. Antoulas, Approximation of Large-Scale Dynamical Systems, Philadelphia, PA, SIAM, 2005.10.1137/1.9780898718713Suche in Google Scholar

[28] P. Benner, S. Gugercin, and K. Willcox, “A survey of projection-based model reduction methods for parametric dynamical systems,” SIAM Review, vol. 57, no. 4, pp. 483–531, 2015, https://doi.org/10.1137/130932715.Suche in Google Scholar

[29] A. Quarteroni, A. Manzoni, and F. Negri, Reduced Basis Methods for Partial Differential Equations: An Introduction, Berlin, Springer, 2016.10.1007/978-3-319-15431-2Suche in Google Scholar

[30] E. J. Parish and K. Duraisamy, “A paradigm for data-driven predictive modeling using field inversion and machine learning,” J. Comput. Phys., vol. 305, pp. 758–774, 2016, https://doi.org/10.1016/j.jcp.2015.11.012.Suche in Google Scholar

[31] B. Recht, “A tour of reinforcement learning: The view from continuous control,” Ann. Rev. Control, Rob., Autonom. Syst., vol. 2, pp. 253–279, 2019, https://doi.org/10.1146/annurev-control-053018-023825.Suche in Google Scholar

[32] M. Fazlyab, A. Robey, H. Hassani, M. Morari, and G. Pappas, “Efficient and accurate estimation of Lipschitz constants for deep neural networks,” Adv. Neural Inf. Process. Syst., vol. 32, 2019.Suche in Google Scholar

[33] M. Revay, R. Wang, and I. R. Manchester, “Recurrent equilibrium networks: Flexible dynamic models with guaranteed stability and robustness,” IEEE Trans. Autom. Control, vol. 69, no. 5, pp. 2855–2870, 2023, https://doi.org/10.1109/tac.2023.3294101.Suche in Google Scholar

[34] S. Tu and B. Recht, “The gap between model-based and model-free methods on the linear quadratic regulator: An asymptotic viewpoint,” in Proceedings of the 32nd Conference on Learning Theory, Arizona, USA, Proceedings of Machine Learning Research (PMLR), 2019, pp. 3036–3083.Suche in Google Scholar

[35] B. Karg and S. Lucia, “Efficient representation and approximation of model predictive control laws via deep learning,” IEEE Trans. Cybern., vol. 50, no. 9, pp. 3866–3878, 2020, https://doi.org/10.1109/tcyb.2020.2999556.Suche in Google Scholar

[36] A. A. Rusu et al.., “Policy distillation,” in Proceedings of the 4th International Conference on Learning Representations, San Juan, Puerto Rico, International Conference on Learning Representations, 2016.Suche in Google Scholar

[37] W. M. Czarnecki, R. Pascanu, S. Osindero, S. Jayakumar, G. Swirszcz, and M. Jaderberg, “Distilling policy distillation,” in Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics, Okinawa, Japan, JMLR (Journal of Machine Learning Research), 2019, pp. 1331–1340.Suche in Google Scholar

[38] M. Tan and Q. Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” in Proceedings of the 36th International Conference on Machine Learning, California, USA, Proceedings of Machine Learning Research (PMLR), 2019, pp. 6105–6114.Suche in Google Scholar

[39] A. Kumar, J. Fu, G. Tucker, and S. Levine, “Data-efficient reinforcement learning with learned controllers,” arXiv:2012.10735, 2020.Suche in Google Scholar

[40] A. Ashok, N. Rhinehart, F. Beainy, and K. M. Kitani, “N2n learning: Network to network compression via policy gradient reinforcement learning,” in Proceedings of the 6th International Conference on Learning Representations, Vancouver, BC, Canada, International Conference on Learning Representations, 2018.Suche in Google Scholar

[41] J. Qiu et al.., “Going deeper with embedded fpga platform for convolutional neural network,” in Proceedings of the 24th ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, Monterey, Association for Computing Machinery, 2016, pp. 26–35.10.1145/2847263.2847265Suche in Google Scholar

[42] A. G. Howard et al.., “Mobilenets: Efficient convolutional neural networks for mobile vision applications,” arXiv:1704.04861, 2017.Suche in Google Scholar

[43] S. Han, H. Mao, and W. J. Dally, “Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding,” in Proceedings of the 4th International Conference on Learning Representations, San Juan, OpenReview, 2016.Suche in Google Scholar

[44] G. Sundaram, J. Ulmen, A. Haider, and D. Görges, “Component-aware pruning for accelerated control tasks in latent space models,” arXiv:2508.08144, 2025.10.1109/SII64115.2026.11404702Suche in Google Scholar

[45] V. Sze, Y.-H. Chen, T.-J. Yang, and J. S. Emer, “Efficient processing of deep neural networks: A tutorial and survey,” Proc. IEEE, vol. 105, no. 12, pp. 2295–2329, 2017, https://doi.org/10.1109/jproc.2017.2761740.Suche in Google Scholar

[46] R. Wilhelm et al.., “The worst-case execution-time problem—overview of methods and survey of tools,” ACM Trans. Embed. Comput. Syst., vol. 7, no. 3, pp. 1–53, 2008, https://doi.org/10.1145/1347375.1347389.Suche in Google Scholar

[47] Y. Kang et al.., “Co-optimizing performance and memory footprint via integrated cpu/gpu memory management, an implementation on autonomous driving platform,” in 2016 IEEE Real-Time and Embedded Technology and Applications Symposium, Vienna, Institute of Electrical and Electronics Engineers (IEEE), 2016, pp. 1–12.Suche in Google Scholar

[48] B. Zoph and Q. V. Le, “Neural architecture search with reinforcement learning,” in Proceedings of the 5th International Conference on Learning Representations, Toulon, OpenReview, 2017.Suche in Google Scholar

[49] J. Mellor, J. Turner, A. Storkey, and E. J. Crowley, “Neural architecture search without training,” in International Conference on Machine Learning, PMLR (Virtual Only Conference), 2021, pp. 7588–7598.Suche in Google Scholar

[50] A. Polino, R. Pascanu, and D. Alistarh, “Model compression via distillation and quantization,” in International Conference on Learning Representations, Vancouver, OpenReview, 2018.Suche in Google Scholar

[51] B. Heo, M. Lee, S. Yun, and J. Y. Choi, “A comprehensive overhaul of feature distillation,” in Proceedings of the 17th IEEE/CVF International Conference on Computer Vision, Seoul, IEEE Computer Society (CPS) / IEEE, 2019, pp. 1921–1930.10.1109/ICCV.2019.00201Suche in Google Scholar

[52] G. Sundaram, J. Ulmen, A. Haider, and D. Görges, “Application-specific component-aware structured pruning of deep neural networks via soft coefficient optimization,” arXiv:2507.14882, 2025.Suche in Google Scholar

[53] S. Teerapittayanon, B. McDanel, and H.-T. Kung, “Branchynet: Fast inference via early exiting from deep neural networks,” in 2016 23rd International Conference on Pattern Recognition, Cancun, IEEE, 2016, pp. 2464–2469.10.1109/ICPR.2016.7900006Suche in Google Scholar

[54] D. Ha and J. Schmidhuber, “World models,” Adv. Neural Inf. Process. Syst., vol. 31, 2018.Suche in Google Scholar

[55] D. Hafner et al.., “Learning latent dynamics for planning from pixels,” in International Conference on Machine Learning, PMLR (Virtual Only Conference), 2019, pp. 2555–2565.Suche in Google Scholar

[56] D. Hafner, T. Lillicrap, M. Norouzi, and J. Ba, “Mastering Atari with discrete world models,” in International Conference on Learning Representations, OpenReview (Virtual Only Conference), 2020.Suche in Google Scholar

[57] D. Hafner, T. Lillicrap, M. Norouzi, and J. Ba, “Mastering atari with discrete world models,” arXiv:2010.02193, 2020.Suche in Google Scholar

[58] K. Asadi, D. Misra, and M. Littman, “Lipschitz continuity in model-based reinforcement learning,” in Proceedings of the 35th International Conference on Machine Learning, Stockholm, PMLR, 2018, pp. 264–273.Suche in Google Scholar

[59] N. Hansen, X. Wang, and H. Su, “Temporal difference learning for model predictive control,” arXiv:2203.04955, 2022.Suche in Google Scholar

[60] N. Hansen, H. Su, and X. Wang, “TD-MPC2: Scalable, robust world models for continuous control,” arXiv:2310.16828, 2024.Suche in Google Scholar

[61] C.-H. Ji, H.-B. Choi, J.-S. Heo, J.-B. Kim, H.-K. Lim, and Y.-H. Han, “Dual policy-based td-learning for model predictive control,” in 2023 International Conference on Artificial Intelligence in Information and Communication, Kuta, Springer Singapore, 2023, pp. 254–258.10.1109/ICAIIC57133.2023.10067037Suche in Google Scholar

[62] H. Lin, P. Wang, J. Schneider, and G. Shi, “TD-M(PC)2: Improving temporal difference MPC through policy constraint,” arXiv:2502.03550, 2025.Suche in Google Scholar

[63] Y. Zhang, S. Yang, T. Ohtsuka, C. Jones, and J. Boedecker, “Latent linear quadratic regulator for robotic control tasks,” arXiv:2407.11107, 2025.Suche in Google Scholar

[64] M. Watter, J. T. Springenberg, J. Boedecker, and M. Riedmiller, “Embed to control: A locally linear latent dynamics model for control from raw images,” arXiv:1506.07365, 2025.Suche in Google Scholar

[65] M. Wang, R. Yang, X. Chen, H. Sun, M. Fang, and G. Montana, “Goplan: Goal-conditioned offline reinforcement learning by planning with learned models,” arXiv:2310.20025, 2023.Suche in Google Scholar

[66] Y. LeCun, “A path towards autonomous machine intelligence version 0.9. 2, 2022-06-27,” Open Review, vol. 62, no. 1, pp. 1–62, 2022.Suche in Google Scholar

[67] M. Assran et al.., “Self-supervised learning from images with a joint-embedding predictive architecture,” arXiv:2301.08243, 2023.10.1109/CVPR52729.2023.01499Suche in Google Scholar

[68] M. Assran et al.., “V-JEPA 2: Self-supervised video models enable understanding, prediction and planning,” arXiv:2506.09985, 2025.Suche in Google Scholar

[69] Z. Fei, M. Fan, and J. Huang, “A-jepa: Joint-embedding predictive architecture can listen,” arXiv:2311.15830, 2024.Suche in Google Scholar

[70] G. Zhou, H. Pan, Y. LeCun, and L. Pinto, “DINO-WM: World models on pre-trained visual features enable zero-shot planning,” arXiv:2411.04983, 2025.Suche in Google Scholar

[71] M. Oquab et al.., “DINOv2: Learning robust visual features without supervision,” Trans. Mach. Learn. Res., pp. 1–31, 2024.Suche in Google Scholar

[72] A. Vujinović and A. Kovačević, “ACT-JEPA: Novel joint-embedding predictive architecture for efficient policy representation learning,” arXiv:2501.14622, 2025.Suche in Google Scholar

[73] V. Sobal, W. Zhang, K. Cho, R. Balestriero, T. G. J. Rudner, and Y. LeCun, “Learning from reward-free offline data: A case for planning with latent dynamics models,” arXiv:2502.14819, 2025.Suche in Google Scholar

[74] T. Kenneweg, P. Kenneweg, and B. Hammer, “JEPA for RL: Investigating joint-embedding predictive architectures for reinforcement learning,” arXiv:2504.16591, 2025.10.14428/esann/2025.ES2025-19Suche in Google Scholar

[75] R. Balestriero and Y. LeCun, “Learning by reconstruction produces uninformative features for perception,” arXiv:2402.11337, 2024.Suche in Google Scholar

[76] E. Littwin et al.., “How JEPA avoids noisy features: The implicit bias of deep linear self distillation networks,” arXiv:2407.03475, 2024.Suche in Google Scholar

[77] S. Mo and S. Tong, “Connecting joint-embedding predictive architecture with contrastive self-supervised learning,” arXiv:2410.19560, 2024.10.52202/079017-0077Suche in Google Scholar

[78] J. Yu et al.., “Why and how auxiliary tasks improve JEPA representations,” arXiv:2509.12249, 2025.Suche in Google Scholar

[79] J. Ulmen, G. Sundaram, and D. Görges, “Learning state-space models of dynamic systems from arbitrary data using joint embedding predictive architectures,” arXiv:2508.10489, 2025.10.1016/j.ifacol.2025.10.190Suche in Google Scholar

[80] M. Hartman and L. R. Varshney, “SparseJEPA: Sparse representation learning of joint embedding predictive architectures,” arXiv:2504.16140, 2025.Suche in Google Scholar

[81] A. Zheng, Evaluating Machine Learning Models: A Beginner’s Guide to Key Concepts and Pitfalls, Sebastopol, CA, O’Reilly Media, 2015.Suche in Google Scholar

[82] J. Eisenstein, Introduction to Natural Language Processing, Cambridge, MA, MIT Press, 2019.Suche in Google Scholar

[83] W. Zhu, “Classification of mnist handwritten digit database using neural network,” Proc. Res. School Comput. Sci. Aust. Natl. Univ., Acton, ACT, vol. 2601, 2018.Suche in Google Scholar

[84] G. Brockman et al.., “OpenAI gym,” arXiv:1606.01540, 2016.Suche in Google Scholar
Received: 2025-10-12
Accepted: 2026-01-09
Published Online: 2026-03-09
Published in Print: 2026-03-26

© 2026 Walter de Gruyter GmbH, Berlin/Boston

Artikel in diesem Heft

  1. Frontmatter
  2. Editorial
  3. 2025 Reisensburg conference on fundamentals of networked and cognitive systems in measurement and automation
  4. Surveys
  5. An overview of sensitivity-based distributed optimization and model predictive control
  6. Deep probabilistic circuits for efficient uncertainty handling in learning systems
  7. Model compression for neural network controllers: a tutorial survey with a focus on controllers with latent state space
  8. Methods
  9. Comparison and performance analysis of dynamic encrypted control approaches
  10. State feedback control for tracking limit cycles of planar periodic systems
https://doi.org/10.1515/auto-2025-0107
Schlagwörter für diesen Artikel
model compression; deep neural networks; latent state space; quantization; pruning; knowledge distillation

Artikel in diesem Heft

  1. Frontmatter
  2. Editorial
  3. 2025 Reisensburg conference on fundamentals of networked and cognitive systems in measurement and automation
  4. Surveys
  5. An overview of sensitivity-based distributed optimization and model predictive control
  6. Deep probabilistic circuits for efficient uncertainty handling in learning systems
  7. Model compression for neural network controllers: a tutorial survey with a focus on controllers with latent state space
  8. Methods
  9. Comparison and performance analysis of dynamic encrypted control approaches
  10. State feedback control for tracking limit cycles of planar periodic systems
Heruntergeladen am 20.4.2026 von https://www.degruyterbrill.com/document/doi/10.1515/auto-2025-0107/html
Button zum nach oben scrollen