When we are working in the phase retrieval configuraiton it is important that each time we re-initialise the data-driven part we use a different seed so that we can restart the optimisation from a different point in the optimisation landscape.
The reset is handled through the TF_NP_poly_OPD.init_vars() method that re-initialises the data-driven part. However, the random seed is set to the config file random seed each time it is done.
If we're running several cycles we need to change the seed that is being set each time we re-initialise the data-driven part.
When we are working in the phase retrieval configuraiton it is important that each time we re-initialise the data-driven part we use a different seed so that we can restart the optimisation from a different point in the optimisation landscape.
The reset is handled through the
TF_NP_poly_OPD.init_vars()method that re-initialises the data-driven part. However, the random seed is set to the config file random seed each time it is done.If we're running several cycles we need to change the seed that is being set each time we re-initialise the data-driven part.