Pradeep Reddy Raamana
Pradeep Reddy Raamana
- Add real world examples - show the effect of different number of ROIs - show the effect of different clustering metrics
- merge from #8 disappeared for some reason :) - dig it or merge it back in
- to unroll 4D data over the 4th dim, - optional input of ROI set or mask - option to cluster the rows (spatial/voxel-wise) and/or columns (time or gradients), -...
- 'middle' returning one middle slice from the views selected - option to select slices within an ROI (`in_roi='seg_image.nii'`)
straight forward implementation of `__add__`, `__mul__`, `__sub__` etc for the `KernelMatrix` class, with some smart optimizations when possible.
so far only full/dense arrays were tested thoroughly. Deeper/broader testing for sparse arrays would be helpful. This is not expected to take much time.
there is existing work on parallelizing KM computation for large N: #3. It needs to be studied thoroughly for computational and storage efficiencies. Also, if the better internal data structure...
this profiling can help find bottlenecks and improve implementation with better internal efficiencies