Skip to content

Package comprising code for "A Sampling Theorem for Deconvolution in Two Dimensions"

Notifications You must be signed in to change notification settings

jpmcd/Deconvolution2D

Repository files navigation

Deconvolution2D

The necessary packages to reconstruct the proof are listed and described below:

File Description
TwoDimBumpWave.wl Library defining functions needed for computational proof.
ServerScript.wl Script computes and saves function tables using Interval Arithmetic, capable of running on a compute cluster.
run-math.sbatch Script to run ServerScript.wl using Slurm scheduler.
segmentdistances.m MATLAB script to compute distances between hexagon cells and segment of length Delta, where Delta is the minimum separation between spikes.
MakeRecoveryTables.wl Script computing monotonized and supremum function envelopes (tables). Pre-computed tables can be loaded. Also computes tables for matrix norms, bump wave coefficients, and recovery results.

To create the function envelopes, the Mathematica script ServerScript.wl must be run. This script computes supremums for bump and wave functions for a selection of grid spacing intervals, defined therein, using Interval Arithmetic. We include a script run-math.sbatch which calls ServerScript.wl and provides options for a cluster managed by Slurm job scheduler.

The MATLAB script segmentdistances.m computes distances between a set of hexagonal cells surrounding the origin and N intervals of equal lengths that partition the segment on the horizontal axis from the origin to (1,0). Its output is a file with name formatted as partition_dists_YYYYMMDD_npartitions_N.mat with date and N substituted. By default we set N=100.

After running segmentdistances.m, recovery results can be obtained by running MakeRecoveryTables.wl. First this script computes envelopes, both monotonized and non-monotonic as needed, for the bump and wave functions using the output of ServerScript.wl. Since this computation is time-consuming these envelopes are saved and can be imported subsequently in future uses by uncommenting the relevant lines. With these envelopes, recovery results are produced for a list of minimum separation sizes.

About

Package comprising code for "A Sampling Theorem for Deconvolution in Two Dimensions"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published