The arithmetic Fourier transform and real neural networks: Summability by primes (Q1892528)

The arithmetic Fourier transform (AFT) is most suitable for the evaluation of Fourier cosine coefficients. The computation is accomplished by use of the Möbius function. The AFT first aroused interest in the field of signal processing because it was suitable for parallel processing. Another feature of the AFT algorithm is that it models the structure of a double layer in an artificial neural network. If the brain (a real neural network) carries out Fourier analysis, is this similar to the AFT rather than conventional Fourier analysis? The purpose of this paper is to present some mathematical results related to the question.