I moved the run_comparison to general.py in utils, then I rewrote the experiment for omega_run_comparison. Of note, I removed the spectral clustering as I found that it is not an effective way of determining similar algorithms. Rather, it is prone to error in a case where a ~ b ~ c, does not mean a ~ c. So, instead, now I just check for pairs of algos that have similarly above a threshold, then increment a count.
I moved the run_comparison to general.py in utils, then I rewrote the experiment for omega_run_comparison. Of note, I removed the spectral clustering as I found that it is not an effective way of determining similar algorithms. Rather, it is prone to error in a case where a ~ b ~ c, does not mean a ~ c. So, instead, now I just check for pairs of algos that have similarly above a threshold, then increment a count.