Abstract
It is a crucial basis to improve the performance of neural network by constructing an appropriate activation function. This paper proposes a novel modulation window radial basis function neural network (MW-RBFNN) with an adjustable activation function. In this MW-RBFNN, a raised cosine radial basis function (RC-RBF) is adaptively modulated by an exponential function, and served as a shape-tunable activation function of MW-RBFNN. Compared with the basic RC-RBF neural network, the approximating ability of MW-RBFNN is improved due to its shape-tunable activation function. Besides, the computation of MW-RBFNN is far less than that of Gaussian radial basis function neural network (GRBFNN) because the MW-RBFNN is compactly supported. The training algorithm of MW-RBFNN is provided and its approximating ability is proved. Moreover, the regulation mechanism of the modulation index for the NN’s performance is proved and the regulating algorithm of the modulation index in MW-RBFNN is given. The computational complexity of MW-RBFNN is also analyzed. Five typical application cases are presented to illustrate the effectiveness of this proposed MW-RBFNN.
Similar content being viewed by others
Availability of data and materials
The datasets used or analysed during the current study are available from the corresponding author on reasonable request.
Code availability
Some or all data, models, or code generated or used are available from the corresponding author by request.
References
Aizenberg I (2010) Periodic activation function and a modified learning algorithm for the multivalued neuron. IEEE Trans Neural Netw 21(12):1939–1949. https://doi.org/10.1109/tnn.2010.2082561
Al-Ajlouni AF, Schilling RJ, Harris SL (2004) Identification of nonlinear discrete-time systems using raised-cosine radial basis function networks. Int J Syst Sci 35(4):211–221. https://doi.org/10.1080/00207720410001703213
Ammar OH, Motai Y (2018) Multicolumn RBF network. IEEE Trans Neural Netw Learn Syst 29(4):766–778. https://doi.org/10.1109/TNNLS.2017.2650865
Apicella A, Isgrò F, Prevete R (2019) A simple and efficient architecture for trainable activation functions. Neurocomputing 370:1–15. https://doi.org/10.1016/j.neucom.2019.08.065
Barros J, Diego RI (2006) On the use of the Hanning window for harmonic analysis in the standard framework. IEEE Trans Power Delivery 21(1):538–539. https://doi.org/10.1109/TPWRD.2005.852339
Chen S, Hong X, Gao JB, Harris CJ (2014) Complex-valued B-spline neural networks for modeling and inverting Hammerstein systems. IEEE Trans Neural Netw Learn Syst 25(9):1673–1685. https://doi.org/10.1109/TNNLS.2014.2298535
Cui Q, Song Y (2021) Tracking control of unknown and constrained nonlinear systems via neural networks with implicit weight and activation learning. IEEE Trans Neural Netw Learn Syst 32(12):5427–5434. https://doi.org/10.1109/TNNLS.2021.3085371
Han ZY, Qian XS, Huang H, Huang TW (2021) Efficient design of multicolumn RBF networks. Neurocomputing 450:253–263. https://doi.org/10.1016/j.neucom.2021.04.040
Hiroshi K, Hirotoshi Y, Toshihiro N (2020) Engineering problems in machine learning systems. Mach Learn 109:1103–1126. https://doi.org/10.1007/s10994-020-05872-w
Huang S, Do B (2014) Radial basis function based neural network for motion detection in dynamic scenes. IEEE Trans Cybern 44(1):114–125. https://doi.org/10.1109/TCYB.2013.2248057
Jafar K, Mehran HP, Javad F (2020) Adaptive neural dynamic surface control of mechanical systems using integral terminal sliding mode. Neurocomputing 379:141–151. https://doi.org/10.1016/j.neucom.2019.10.046
Jafar K, Hadi MK, Peiman N (2022) Adaptive neural dynamic surface control for uniform energy exploitation of floating wind turbine. Appl Energy 316:119132. https://doi.org/10.1016/j.apenergy.2022.119132
Lei Y, Ding L, Zhang W (2015) Generalization performance of radial basis function networks. IEEE Trans Neural Netw Learn Syst 26(3):551–564. https://doi.org/10.1109/TNNLS.2014.2320280
Li B, Li Y, Rong X (2013) The extreme learning machine learning algorithm with tunable activation function. Neural Comput Appl 22(3–4):531–539. https://doi.org/10.1007/s00521-012-0858-9
Lian J, Lee Y, Sudhoff SD, Zak SH (2008a) Self-organizing radial basis function network for real-time approximation of continuous-time dynamical systems. IEEE Trans Neural Netw 19(3):460–474. https://doi.org/10.1109/TNN.2007.909842
Lian J, Lee Y, Sudhoff SD, Zak SH (2008b). Variable structure neural network based direct adaptive robust control of uncertain systems. In: American Control Conference, Washington, USA. pp. 3402–3407. https://doi.org/10.1109/ACC.2008.4587018
Lian J, Hu J, Zak SH (2015) Variable neural adaptive robust control: a switched system approach. IEEE Trans Neural Netw Learn Syst 26(5):903–915. https://doi.org/10.1109/TNNLS.2014.2327853
Liu Y, Tang L, Tong S, Chen CP (2015) Adaptive NN controller design for a class of nonlinear MIMO discrete-time systems. IEEE Trans Neural Netw Learn Syst 26(5):1007–1018. https://doi.org/10.1109/TNNLS.2014.2330336
Liu W, Wang Z, Yuan Y, Zeng N, Hone K, Liu X (2021) A novel sigmoid-function-based adaptive weighted particle swarm optimizer. IEEE Trans Cybern 51(2):1085–1093. https://doi.org/10.1109/TCYB.2019.2925015
Liu YB, Sun WC, Gao HJ (2022) High precision robust control for periodic tasks of linear motor via B-spline wavelet neural network observer. IEEE Trans Ind Electron 69(8):8255–8263. https://doi.org/10.1109/TIE.2021.3102426
Liu Z, Leung CS, So HC (2023) Formal convergence analysis on deterministic ℓ1-regularization based mini-batch learning for RBF networks. Neurocomputing 532:77–93. https://doi.org/10.1016/j.neucom.2023.02.012
Luo S, Li S, Tajaddodianfar F, Hu J (2018) Adaptive synchronization of the fractional-order chaotic arch micro-electro-mechanical system via Chebyshev neural network. IEEE Sens J 18(9):3524–3532. https://doi.org/10.1109/JSEN.2018.2812859
Mahdi RN, Rouchka ER (2011) Reduced hyperBF networks: regularization by explicit complexity reduction and scaled rprop-based training. IEEE Trans Neural Netw 22(5):673–686. https://doi.org/10.1109/TNN.2011.2109736
Miao P, Shen Y, Xia X (2014) Finite time dual neural networks with a tunable activation function for solving quadratic programming problems and its application. Neurocomputing 143(S1):80–89. https://doi.org/10.1016/j.neucom.2014.06.018
Min H, Xu S, Fei S, Yu X (2021) Observer-based NN control for nonlinear systems with full-state constraints and external disturbances. IEEE Trans Neural Netw Learn Syst 33(9):4322–4331. https://doi.org/10.1109/TNNLS.2021.3056524
Pan Y, Yu H, Er MJ (2014) Adaptive neural PD control with semiglobal asymptotic stabilization guarantee. IEEE Trans Neural Netw Learn Syst 25(12):2264–2274. https://doi.org/10.1109/TNNLS.2014.2308571
Patra JC, Kot AC (2002) Nonlinear dynamic system identification using Chebyshev function link artificial neural networks. IEEE Trans Syst, Man Cybern: B, Cybern 32(4):505–511. https://doi.org/10.1109/TSMCB.2002.1018769
Prakash C, Alka S, Madhusudan S (2018) Chebyshev functional expansion based artificial neural network controller for shunt compensation. IEEE Trans Ind Inform 14(9):3792–3800. https://doi.org/10.1109/TII.2018.2793347
Qian S, Liu H, Liu C, Wu S, Wong HS (2018) Adaptive activation functions in convolutional neural networks. Neurocomputing 272:204–212. https://doi.org/10.1016/j.neucom.2017.06.070
Raitoharju J, Kiranyaz S, Gabbouj M (2016) Training radial basis function neural networks for classification via class-specific clustering. IEEE Trans Neural Netw Learn Syst 27(12):2458–2470. https://doi.org/10.1109/TNNLS.2015.2497286
Sabato MS, Valerio MS (2017) Adaptation to new microphones using artificial neural networks with trainable activation functions. IEEE Trans Neural Netw Learn Syst 28(8):1959–1965. https://doi.org/10.1109/TNNLS.2016.2550532
Sahoo DM, Chakraverty S (2018) Functional link neural network learning for response prediction of tall shear buildings with respect to earthquake data. IEEE Trans Syst, Man, Cybern: Syst 48(1):1–10. https://doi.org/10.1109/TSMC.2017.2700334
Schilling RJ, Carroll JJ, Al-Ajlouni AF (2001) Approximation of nonlinear systems with radial basis function neural networks. IEEE Trans Neural Netw 12(1):1–15. https://doi.org/10.1109/72.896792
Segee BE (1993) Using spectral techniques for improved performance in artificial neural networks. In: IEEE International Conference on Neural Networks, USA, pp. 500–505. https://doi.org/10.1109/ICNN.1993.298608
Tan Z, Hu Y, Chen K (2020) On the investigation of activation functions in gradient neural network for online solving linear matrix equation. Neurocomputing 413:185–192. https://doi.org/10.1016/j.neucom.2020.06.097
Uykan Z (2013) Fast-convergent double-sigmoid hopfield neural network as applied to optimization problems. IEEE Trans Neural Netw Learn Syst 24(6):990–996. https://doi.org/10.1109/TNNLS.2013.2248092
Vyas BY, Das B, Maheshwari RP (2016) Improved fault classification in series compensated transmission line-comparative evaluation of Chebyshev neural network. IEEE Trans Neural Netw Learn Syst 27(8):1631–1642. https://doi.org/10.1109/TNNLS.2014.2360879
Wang D, Zeng JS, Lin SB (2021) Random sketching for neural networks with ReLU. IEEE Trans Neural Netw Learn Syst 32(2):748–762. https://doi.org/10.1109/TNNLS.2020.2979228
Webb A, Shannon S (1998) Shape-adaptive radial basis functions. IEEE Trans Neural Networks 9(6):1155–1166. https://doi.org/10.1109/72.728359
Wen X, Zhang X, Zhu Y (2015) Design of fault detection observer based on hyper basis function. Tsinghua Sci Technol 20(2):200–204. https://doi.org/10.1109/TST.2015.7085633
Wu Y, Zhao M, Ding X (1996) A new perceptron model with tunable activation function. Acta Electron Sin 5(2):55–62
Yang QL, Sadeghi A, Wang G, Sun J (2021) Learning two-layer ReLU networks is nearly as easy as learning linear classifiers on separable data. IEEE Trans Signal Process 69(7):4416–4427. https://doi.org/10.1109/TSP.2021.3094911
Yin KL, Pu YF, Lu L (2020) Hermite functional link artificial-neural-network-assisted adaptive algorithms for IoV nonlinear active noise control. IEEE Internet Things J 7(9):8372–8383. https://doi.org/10.1109/JIOT.2020.2989761
Yu H, Reiner PD, Xie T, Bartczak T, Wilamowski BM (2014) An incremental design of radial basis function networks. IEEE Trans Neural Netw Learn Syst 25(10):1793–1802. https://doi.org/10.1109/TNNLS.2013.2295813
Zhang H, Zeng Z (2021) Synchronization of nonidentical neural networks with unknown parameters and diffusion effects via robust adaptive control techniques. IEEE Trans Cybern 51(2):660–667. https://doi.org/10.1109/TCYB.2019.2921633
Zhang Z, Zhang J (2022) Chebyshev functional link spline neural filter for nonlinear dynamic system identification. IEEE Trans Circuits Syst II: Express Briefs 69(3):1907–1911. https://doi.org/10.1109/TCSII.2021.3111919
Zhang X, Han Q, Zeng Z (2018) Hierarchical type stability criteria for delayed neural networks via canonical Bessel-Legendre Inequalities. IEEE Trans Cybern 48(5):1660–1671. https://doi.org/10.1109/TCYB.2017.2776283
Zhao Y, Pei J, Chen H (2019) Multi-layer radial basis function neural network based on multi-scale kernel learning. Appl Soft Comput 82:105541. https://doi.org/10.1016/j.asoc.2019.105541
Acknowledgements
The authors would like to appreciate the National Natural Science Foundation of China (grant number 51775185), and the Natural Science Foundation of Hunan Province, China (grant number 2018JJ2261), for their kind supports.
Funding
This study was supported in part by grants from the National Natural Science Foundation of China (grant number 51775185), and the Natural Science Foundation of Hunan Province, China (grant number 2018JJ2261).
Author information
Authors and Affiliations
Contributions
HL proposed the MW-RBFNN and gave the training algorithm of MW-RBFNN, and was a major contributor in writing the manuscript. HD prove the approximating ability of MW-RBFNN, and analyzed the computation of MW-RBFNN. LW studied the experiments. YM designed the experimental code. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interests.
Ethical approval
Not Applicable.
Consent to participate
Not Applicable.
Consent for publication
Not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix I
Appendix I
MW-RBFNN-based Jacobian in PID control system.
The output of the MW-RBFNN shown in Fig. 8 can be rewritten as
where x = [Δu(k), y(k), y(k-1)]T, and \(\phi_{Wk} ({\mathbf{x}})\) is descried by (10). Therefore,
where,
where x1 = Δu(k), pi(xi) is described by (5), and
Thus, in (45), \(\frac{{\partial \zeta_{c} \left[ {p_{1} (x_{1} ) - p_{1} (x_{1}^{(k)} )} \right]}}{\partial \Delta u}\) is obtained by (47), i.e.,
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Lin, H., Dai, H., Mao, Y. et al. An optimized radial basis function neural network with modulation-window activation function. Soft Comput 28, 4631–4648 (2024). https://doi.org/10.1007/s00500-023-09207-4
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00500-023-09207-4