Version 23.1.0
Install using files from https://anaconda.org/ccpi/cil/files?version=23.1.0
v23.1.0
The full changelog can be found here.
This version extends the TotalVariation class to store its state via a warm_start
flag and changes the default behaviour to use this.
This means that calls to TotalVariation::proximal()
are initiated by the final value from the previous iteration which allows the max_iteration
value to be reduced significantly. When used within an algorithm such as PDHG the total memory use will increase by the number of dimensions times the image size, however the TotalVariation
step will need only 5-10 iterations, significantly decreasing overall compute time until convergence. The previous behaviour can be recreated by setting warm_start=False
.