Margaret Duff

icon company STFC - UKRI icon location Computational Scientist in Imaging - STFC - UKRI

Results 90 comments of Margaret Duff

Algorithm should be able to compute "sublosses" efficiently on demand

Thanks @manchester-jhellier - I think this is complex, because for some algorithms e.g. PDHG we might not ever evaluate the individual functions on the current iterate during the update. However,...

Sqlite Callback

Hey @manchester-jhellier - thank-you for this - it looks really interesting! A demonstration example could be something on the spheres data (which is small and *should* be quick to run)...

Sqlite Callback

Thanks @manchester-jhellier!

DataContainer should probably have a save method

Thanks @manchester-jhellier - This is something we have talked about a lot recently. Our sister package SIRF, does indeed have a save method for data containers. Definitely something we need...

Power method tolerance 1e-5 should be smaller

We will investigate this but in the meantime we should at least change the warning

Tigre algorithm wrapping

Discussed with developers today: - We would like to try a general wrapping of tigre algorithms e.g. tigre.algo( name, initial, image_geometry, data, niter, **kwargs) which then converts the geometry, data...

Tigre algorithm wrapping

Discussed with developers: - We prefer option 2, it works with all current tigre algorithms and it is clearer that it is Tigre's algorithm and the credit goes to them...

Tigre algorithm wrapping

@gfardell - I was having another think about this - we discussed writing the wrapper as a processor but to force the code to have an `init`, `set_input` and `get_output`...

Remove deprecated code for v25.0.0

Thanks @lauramurgatroyd - I am happy with those changes if it passes all the tests.

BlockOperator direct and adjoint methods: can pass out as a DataContainer instead of a (1,1) BlockDataContainer where geometry permits

So in #1802 we changed BlockOperator direct (and adjoint) methods to return a DataContainer in the case it would previously return a (1,1) BlockDataContainer. We also changed the range (and...