GitLab

Changed DSS to include the difficulty as intended. FUTURE ISSUE: Race-conditions there at increasing the difficulty, when parallel.

Simplied DSS was implemented. Works alright, but not a huge benefit over the random sampling in areas.

Added limit-checking to islands, and better output after it's done by combining the sub-island-genomes into 1 big one as a set (unique individuals).

Added a lil TODO. For very important things, using gitlab and adding some milestones might be applicable. But here can be some ideas be added, or general things

Still takes 8-10 minutes, but produces good results relyably!

Example-run was done in ~8 minutes, first solution found after around 5.

Started working on the island-model.

Now creates file with the pids of running programs, and all can be killed by evoking a single bashscript.

Changed fitness to include #unique_positions.

TODO: Insert that as the crossover-method into the genetic program.

Made the fitness depend on the distance to goal after the execution (not during), and working programs now have the #iterations, #steps and #loc as fitness (As sum. Called pareto-front, but i didnt look much into that yet. Might need different coefficients, or depending on previous values in the formula.)

Distance now changes after every loc, instead of every iteration. This will punish wrong moves, and praise good ones.
Also, added little shellscripts to mount the output-folder into RAM.

Distance now changes after every loc, instead of every iteration. This will punish wrong moves, and praise good ones.

Better: Generate when run starts.

Also, checked if the test for the correct prog still works on random samples, which it does.