Nonlinear activation functions for machine learning via SHG on PPTFLN

We explore the potential of second harmonic generation (SHG) on periodically poled thin-film lithium niobate (PPTFLN) to enhance nonlinear activation functions in machine learning applications. By leveraging the unique nonlinear optical properties of PPTFLN, SHG can introduce efficient, high-speed transformations essential for advanced neural network architectures. Initial results demonstrate successful SHG in PPTFLN, laying the groundwork for future integration with machine learning systems. While direct application to machine learning is still under development, this approach promises a novel hardware-based activation mechanism, potentially reducing computational demands and power consumption compared to traditional activation functions in artificial neural networks.