Title : Behavior of bessel function as activation function in neuron with multi dendrites
Abstract:
This study examines the use of Bessel functions as activation functions within multi-dendritic neurons. Activation functions are fundamental components of neural networks, introducing crucial non-linearity for learning complex patterns in data. Traditional activation functions such as sigmoid and ReLU have specific characteristics, but their suitability for more complex neural architectures remains underexplored. Bessel functions offer unique mathematical properties suitable for this analysis. They exhibit varying degrees of non-linearity based on their order and arguments, making them intriguing candidates for shaping neural responses. By incorporating Bessel functions into neurons with multiple dendrites, we aim to understand their impact on neural network dynamics, learning capabilities, and computational efficiency. Our findings will highlight how different orders and parameters of Bessel functions influence network convergence, learning speed, and generalization. This research contributes to the intersection of mathematical functions and neural computation, providing insights into innovative approaches for designing activation functions tailored to complex neural architectures. The results will inform advancements in neural network theory and optimization, with implications for both artificial and biological neural systems.
Keywords: Neural Networks, Bessel Function, Activation Function, Neuron, Multi Dendrites.
