This is the companion repository for the BSc project "Randomised Methods for Computing the QR Decomposition of a Tall-Skinny Matrix" by Artis Vijups.
A spec file spec.yml is provided that can be used to recreate the conda environment.
There are multiple .sh scripts provided. Specify account/partition inside all five .sh files and write:
sbatch <script>to execute one of the scripts described by the table below. The scripts with corresponding notebooks generate .csv files as results, and the notebooks generate tables or .png plots.
| Script | Purpose | Observed duration | Notes | Corresponding notebook |
|---|---|---|---|---|
h5create.sh |
Create test matrix sets | 2-3 minutes | Must be run first. Ensure 30 GB of available local storage. | - |
accuracy.sh |
Run accuracy tests | 2-3 minutes | - | accuracy.ipynb |
runtime.sh |
Run runtime tests | 20-25 minutes | Reserves three nodes instead of one. Creates seven jobs of runtimehelper.sh which call runtimehelper.py. |
runtime.ipynb |
sample.sh |
Run sampling rate tests | 7-8 minutes | - | sample.ipynb |