A team of Rice University researchers trained a deep neural network using Fourier analysis, a 200-year-old mathematical tool, to quickly train more accurate neural networks. Fourier analysis provides a way to decompose complex waves into simpler components, making it possible to study the behavior of the networks millions of parameters. The research team believes that this approach can provide new ideas on how to open up the black box of artificial intelligence and make these systems more transparent and understandable. IEEE Spectrum is the flagship publication of the IEEE, the worlds largest professional organization devoted to engineering and applied sciences.
