Abstract:According to harmonic analysis (Fourier analysis), any function $f(x)$, periodic over the interval $[-L, L]$, which satisfies the Dirichlet conditions, can be developed into an infinite sum (known in the literature as the trigonometric series, and for which, for reasons which will become evident in the course of this work, we will use the name of sinusoidal series), consisting of the weighted components of a complete biortogonal base, formed of the unitary function 1, of the fundamental harmonics $\sin(\pi x/L)$, even and $\cos(\pi x/L)$, odd ($2L$-periodic functions) and of the secondary harmonics $\sin(n\pi x/L)$ and $\cos(n\pi x/L)$ (periodic functions, with period $2L/n$, where $n\in \mathbb{Z}^+$, positive integers). The coefficients of these expansions (Fourier coefficients) can be calculated using Euler formulas. We will generalize this statement and show that the function $f(x)$ can also be developed into non-sinusoidal periodic series, formed from the sum of the weighted components of a complete, non-orthogonal base: the unit function 1, the fundamental quasi-harmonics $g(x)$, even and $h(x)$, odd ($2L$-periodic functions, with zero mean value over the definition interval) and the secondary quasi-harmonics $g_n(x)$ and $h_n(x)$ ($2L/n$-periodic functions), where $n\in \mathbb{Z}^+$. The fundamental quasi-harmonics $g(x)$ and $h(x)$ are any functions which admit expansions in sinusoidal series (satisfy Dirichlet conditions, or belong to $L^2$ space). The coefficients of these expansions are obtained with the help of certain algebraic relationships between the Fourier coefficients of the expansions of the functions $f(x)$, $g(x)$ and $h(x)$. In addition to their obvious theoretical importance, these types of expansions can have practical importance in the approximation of functions and in the numerical and analytical resolution of certain classes of differential equations.

Taming the Frequency Factory of Sinusoidal Networks

Understanding Sinusoidal Neural Networks

Beyond Periodicity: Towards a Unifying Framework for Activations in Coordinate-MLPs

The Neural Network Model of Sinusoid Activation Transfer Function

Fast Training of Sinusoidal Neural Fields via Scaling Initialization

Data-driven Estimation of Sinusoid Frequencies

Implicit Neural Representation of Tileable Material Textures

Implicit Neural Representations with Periodic Activation Functions

Learning High-Frequency Functions Made Easy with Sinusoidal Positional Encoding

An MP-based Harmonic and Individual Lines Sinusoidal Modeling in Parametric Audio Coding

Implicit Neural Representations and the Algebra of Complex Wavelets

Simo Fourier Neural Networks Research

Amplifying Sine Unit: An Oscillatory Activation Function for Deep Neural Networks to Recover Nonlinear Oscillations Efficiently

FAN: Fourier Analysis Networks

Understanding the dynamics of the frequency bias in neural networks

Nonsinusoidal periodic Fourier series

A Neural Network for Parameter Estimation of the Exponentially Damped Sinusoids.

Training via quantum superposition circumventing local minima and vanishing gradient of sinusoidal neural network

Sparse DNN Model for Frequency Expanding of Higher Order Ambisonics Encoding Process

Line Spectral Estimation Inspired by Quasi-Neural Network

On Optimal Sampling for Learning SDF Using MLPs Equipped with Positional Encoding