006 super resolution

[chaozg]

fixed #26

• [x] implement smoothed Laplace
• [x] reduce dimension through an likelihood approximation
• [x] CUQIpy-Pytorch's NUTS

This notebook shows a combined use of smoothed Laplace prior, reduced likelihood computation and efficient gradient sampler;

• The forward model and the associated indices for the reduced likelihood are imported from binary files prepared by the author;
• This notebook uses NUTS from CUQIpy-PyTorch;
• The sampling is rather expensive, consuming like 20 hours on the gbar HPC;
• The index of this notebook will be 12 as 6 is taken by the one on PnP

[chaozg]

It looks great and I like the additional final explanations added as agreed this morning.

It seems the generated outputs were not saved in this version, so please can you rerun and include the outputs, so the notebooks can be browsed with results directly on github.

Also, in the readme, please add in the list of topics covered: CUQIpy-pyTorch plugin.

Done, sorry for not noticing the loss of outputs (it seems due to an accident merge) but they are back now.

[chaozg]

Thank @jakobsj and @raff0722 for your review. I'm now merging this branch to the main.