Higher-Order Deep Polynomial Neural Network

MATLAB implementation of a Deep Polynomial Neural Network, a higher-order architecture where neurons compute polynomial combinations of inputs (x, x², x³, ..., xᴰ) rather than simple weighted sums. This design enables the network to inherently capture complex non-linear relationships. The code includes the polynomial layer implementation, training algorithms, and Bayesian optimization for hyperparameter tuning (polynomial degree, network depth, layer sizes). Note: The paper is submitted to ICCAD 2026. Restrictions will be removed upon paper acceptance.