Publication of Astronomical Observatory of Belgrade, Vol. 105, 2024, Page 6, https://doi.org/10.69646/14sbac05a
BOOK OF ABSTRACTS: XIV Serbian-Bulgarian Astronomical Conference, 23 - 27 September, 2024, Vrnjačka Banja, Serbia. Editors: Milan S. Dimitrijević, Evgeni Semkov, Zoran Simić, Goran Damljanović, Momchil Dechev

We study some new approaches to Neural networks based on spline versions of the Kolmogorov-Arnold Networks, which have been introduced and studied recently. We explore the flexibility of this new approach, for solving some particular problems of Machine learning and AI. A main novelty in our approach is that we explore KANs based on multivariate polysplines.

 

References:

B. Igelnik and N. Parikh. Kolmogorov's spline network. IEEE Transactions on Neural Networks, 14(4):725--733, 2003.

Johannes Schmidt-Hieber, The Kolmogorov-Arnold representation theorem revisited, Neural Networks, Volume 137, May 2021, Pages 119-126.

Andrew Polar, Michael Poluektov, A deep machine learning algorithm for construction of the Kolmogorov-Arnold representation. Engineering Applications of Artificial Intelligence,Volume 99, March 2021, 104137

O. Kounchev, Multivariate Polsyplines. Applications to Numerical and Wavelet Analysis, Academic Press-Elsevier, 2001.