prottonicfusion / parallel-genetic-optimization Goto Github PK

If the i th gene of both parents is already contained in the child route, a replacement should be chosen in a sensible way

Currently shared with all 4 neighbors, which probably unnecessary and wasteful

Especially when run on a small number of CPUs, 'spikes' of worsening results can be seen

A 15 city system should probably be optimized with under 100 generations, using a population of 50-100

Verify that the file indeed contains as many cities as stated on the first line etc.

The class should contain at least:

• the population size
• the individuals making up the population
• functions for:
  • initializing a new random population
  • returning the fittest individual

The overhead introduced by MPI communication should be reduced.
Already checked:

• Inter-processor migration of individuals
• Output to screen & file; changing frequency has virtually no impact

Exit main calculation loop if bestRoute does not change over a long time