Dataset

Code to reconstruction transmission matrices from reflection matrices, as well as some sample matrices used to produce Figure 8 of the paper.

Instructions: Run the file localoptimisation.py using Python. You will need to have Tensorflow installed (I recommend the GPU version). Run the code once until it starts to oscilliate then run it again (starting from the previous starting point) with a different optimiser, e.g. basic Stochastic Gradient instead of ADAM, or with a lower learning rate. That often helps to get the recovered matrix very close to the ideal. The code first generations reflection matrices and then uses the known reflectors and wavelengths to recover the original transmission matrices. Files: currentGraph_sample.png shows the type of response graph you might expect if starting from a random starting point. reflectors.mat give the reflector matrices used simSettings.mat gives the wavlengths etc. used in the simulation target.mat gives the multi-wavelength matrices you are trying to recover