Receiving device and method for determining transmission characteristics of an optical waveguide

A method for determining transmission characteristics of an optical waveguide, the method comprising:

• receiving a blended light beam from the optical waveguide, the blended light beam comprising an at least partially superimposed plurality of light beams, each light beam of the plurality of light beams being associated with a mode of a plurality of modes with an associated amplitude and an associated phase;
• determining, using a trained neural network, mode information for the received intermixed light beam, the mode information comprising the associated amplitude and associated phase information for each mode of the plurality of modes;
• receiving a mixed shifted light beam from the optical waveguide, the mixed shifted light beam comprising an at least partially superimposed plurality of shifted light beams, each shifted light beam of the plurality of shifted light beams being associated with a light beam of the plurality of light beams, the shifted light beam associated with a respective light beam having the associated mode with the associated amplitude and an associated shifted phase, the shifted phase being different from the phase of the light beam;
• determining, using the trained neural network, shifted mode information for the intermixed shifted light beam, the shifted mode information comprising, for each mode of the plurality of modes, the associated amplitude and associated shifted phase information;
• and
• determining, for each mode of the plurality of modes, the respective associated phase of the light beam using the associated phase information of the light beam and the associated shifted phase information of the shifted light beam.

A method of computer-based training of a neural network, comprising:

• providing a plurality of first amplitude images, each first amplitude image of the plurality of first amplitude images having associated therewith a first plurality of modes, wherein, for each first amplitude image of the plurality of first amplitude images, each mode of the first plurality of modes has associated therewith a respective amplitude and a respective phase;
• determining, using the neural network, mode information for each first amplitude image of the plurality of first amplitude images, the mode information comprising, for each mode of the first plurality of modes, a determined associated amplitude and a determined associated phase information;
• pre-training the neural network, for each first amplitude image of the plurality of first amplitude images, using the amplitudes provided and phases provided for the first plurality of modes and the amplitudes determined and phase information determined for the first plurality of modes;
• providing a plurality of second amplitude images, each second amplitude image of the plurality of second amplitude images being associated with a second plurality of modes, the second plurality of modes comprising the first plurality of modes and at least one additional mode, wherein, for each second amplitude image of the plurality of second amplitude images, each mode of the second plurality of modes has associated therewith a respective amplitude and a respective phase;
• determining, using the neural network, mode information for each second amplitude image of the plurality of second amplitude images, the mode information for each mode of the second plurality of modes having a determined associated amplitude and determined associated phase information;
• and
• training the neural network, for each second amplitude image of the plurality of second amplitude images, using the amplitudes provided and phases provided for the second plurality of modes and the amplitudes determined and phase information determined for the second plurality of modes.