<ul data-eligibleForWebStory="true">Exploring complex-valued neural networks reveals hidden periodic alter-egos of activation functions.Transitioning neural networks to complex numbers introduces challenges with activation function differentiability.Complex-Valued Neural Networks (CVNNs) show promise in wave-related fields but face activation function limitations.Traditional activation functions like tanh become singular and non-differentiable in the complex plane.There is a trade-off between boundedness and differentiability leading to need for specialized functions.