Constructions in Combinatorics via Neural Networks

Code accompanying the bachelor thesis "Constructions in Combinatorics via Neural Networks", written by Colin Doumont and supervised by Lars Rohwedder.

Software requirements

See requirements.txt file for necessary Python packages.

Files and usage

1. basic.py: basic adaption of Wagner's code for discrepancy theory
2. parallel.py: identical to basic.py, except it runs in parallel on all available CPUs
3. fractional.py: more sophisticated adaptation of Wagner's code, now also working for fractional discrepancy
4. discrepancy.py: discrepacy-related functions used in files 1, 2 and 3
5. dynamic.py: dynamic program used in files 1 and 2
6. random_search.py: naive approach to finding matrices with a certain discrepancy
7. NN_search.py: identical to parallel.py, except it runs for 100 trials and then logs the average
8. plotting.py: plots for the figures in the paper