I worry about the amount of copy/pasting I had to do in the init function of the MaximumAPosteriori class in bayeseor/utils/map.py. I think it would be better to modularize the noise and k cube generation a bit more, but those changes felt outside the scope of this branch.
The MaximumAPosteriori class has been tested and can reliably produce maximum a posteriori visibilities using GPUs.
Changes introduced in this merge:
Added a MaximumAPosteriori class to bayeseor/utils/map.py to calculate maximum a posteriori estimates given an input power spectrum
Added a gpu submodule bayeseor.gpu which contains a new class GPUInterface to interface with GPUs
I worry about the amount of copy/pasting I had to do in the init function of the MaximumAPosteriori class in
bayeseor/utils/map.py. I think it would be better to modularize the noise and k cube generation a bit more, but those changes felt outside the scope of this branch.The MaximumAPosteriori class has been tested and can reliably produce maximum a posteriori visibilities using GPUs.
Changes introduced in this merge:
bayeseor/utils/map.pyto calculate maximum a posteriori estimates given an input power spectrumbayeseor.gpuwhich contains a new classGPUInterfaceto interface with GPUs