Neural networks for the compensation of nonlinear noise influences on the nonlinear Fourier transformation
The modelling of the propagation of light waves through an optical fiber is not trivial. Due to nonlinear effects in the fiber, the evolution of the light wave depends, among other things, on its instantaneous power. In order to increase the data rate in today's systems, it is necessary to choose the signal power as high as possible, since this maximizes the ratio of signal power to noise power. However, due to the nonlinear effects in the optical fiber, additional interference gradually occurs at high power levels, limiting the maximum power and thus the maximum data rate.
A mathematical tool to minimize these nonlinear interferences is the socalled nonlinear Fourier Transform (NFT). The NFT takes into account the nonlinear interferences in the fiber and modulates the signals to be transmitted accordingly. In this way, the maximum usable transmission power can be increased and, as a result, the maximum data rate can be increased.
At this point, neural networks can be used to learn these nonlinear noise influences and to compensate them as good as possible. In this way the maximum transmission range and the data rate can be increased.
A mathematical tool to minimize these nonlinear interferences is the socalled nonlinear Fourier Transform (NFT). The NFT takes into account the nonlinear interferences in the fiber and modulates the signals to be transmitted accordingly. In this way, the maximum usable transmission power can be increased and, as a result, the maximum data rate can be increased.
At this point, neural networks can be used to learn these nonlinear noise influences and to compensate them as good as possible. In this way the maximum transmission range and the data rate can be increased.
funded by
Publications:
