Evolutionary Algorithms Project (Group ??)

CS4205 Evolutioanry Algorithms members :

• Sharwin Bobde (5011639)
• Thalis Papakyriakou
• Isha Dijcks
• Rickard Hellström

Instructions

save the maxcut instances under /data/maxcut. Example of a .txt file relative path would be data/maxcut/set0a/n0000006i00.txt

Methodology

Finding n_req Scalability analysis

• Let problem size be number of vertices $v$
• number of $v$ can be obrained from the static method GraphManager.get_graph_files()

Finding the population $n_{req}$ required to solve a problem 10/10 times

• find $n_{upper}$ using n = [2, 4, 6 , 8 ...]
• Let $n_{lower} = \frac{n_{upper}}{2}$
• Search $n_{req}$ between $n_{lower}$ and $n_{upper}$ by increasing $n$ by a factor of 1.1 each time

Model Comparisons

We will be doing the comparison for both Black-box and Grey/White-box approach. Save all observations in csv as following:

• filename <EA-used>-<B or W><metric name>-<set_name>.csv
• with the columns v, <metric name>_mean, <metric name>_std
• Examples:
  • particle_swarm-B-num_eval-set0b.csv with columns v, num_eval_mean, num_eval_std
  • particle_swarm-W-runtime-set0a.csv with columns v, runtime_mean, runtime_std
  • particle_swarm-B-gen-set0c.csv with columns v, gen_mean, gen_std

Metrics

• number of fitness evaluations (mean and std. for 10 runs of 5 graphs per problem set) (num_eval)
• runtime in sec. (mean and std. for 10 runs of 5 graphs per problem set) (runtime)
• Np. of generations. (mean and std. for 10 runs of 5 graphs per problem set) (gen)