Major speed and memory improvements
This release improves the speed performance of the code and introduces tools for interactive visualization. Major changes include:
- Use of the CVXPY library together with the ECOS and SCS solvers for fast updating of the temporal components. To use this solver set
options['temporal_params']['method'] = 'cvxpy'
. Installation of the CVXPY library is required. - Manual component refinement through
utilities.manually_refine_components
. Allows the user to select additional components by clicking on a summary statistic image. A new component is then fit around that neighborhood. - Setting the AR order
p=0
throughoptions['temporal_params']['p'] = 0
turns off deconvolution. Use this for fast updating of temporal components during first iteration and merging for speed improvements. - Memory mapping for large data handling using little RAM. In the demo data is reshaped into a 2d array, then saved as numpy array that is used in a memory mapped form.
- Fixed a bug in
utilities.order_components
update_temporal_components
also outputs the component-filtered noise statistic YrA. This can be used as an input toutilities.view_patches
andutilities.view_patches_bar
for faster visualization.- General cleanup and code optimization.