Different complexities will be obtained as the matrix numbers will change according to different network structures (hidden layer etc.) and learning algorithms. There is a detailed explanation on the link: https://medium.com/swlh/computational-complexity-of-neural-networks-38c01e7e566a

I suggest you follow https://openreview.net/pdf?id=ByxkijC5FQ and https://arxiv.org/ftp/arxiv/papers/1810/1810.11760.pdf

Preprint Neural Complexity Measures

Preprint Measuring Model Complexity of Neural Networks with Curve Act...

Maybe just count model size.

I think you can use the information measures of the complexity of the systems that I proposed: http://lc.kubagro.ru/aidos/Work_on_emergence.htm

The more the hidden layer, inputs and outputs are, the more the complexity.

I proposed:

Neural Complexity Measures

Yoonho Lee, Juho Lee, Sung Ju Hwang, Eunho Yang, Seungjin Choi

Measuring Model Complexity of Neural Networks with Curve Activation Functions

Xia Hu, Weiqing Liu, Jiang Bian, Jian Pei

Computational Complexity of Neural Networks

On why neural networks are so slow to train and how can we speed it up

Mohammad Ahmad

https://openreview.net/pdf?id=ByxkijC5FQ

La complexité d'un réseau de neurones se resoud par l'introduction de l'algorithme d'apprentissage back propagation et aussi d'un gradient of errors qui améliorent la précision de la

mesure. Je vous suggère mon travail:https:://www.researchgate.net/publication/347901744_Final_dissertation_Andre_Mutunda