Fitness keeps jumping up in new, faster version of code.

[Treblesteph]

Before (slow code from master branch) the fitness seemed to be monotonically decreasing. Now (fast code in speedup branch) there are sudden jumps where lower fitnesses creep into first place by somehow out-competing the fittest members:

generation 1696. mean fitness: 0.28. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634}xxxxx.l...l..l.l..p.p..p...l......l..l....l....l.p...e.g...........p.e..
generation 1697. mean fitness: 0.31. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634}xxxx......e..p..g..l..l....l.......l......l...l......l...l.......g
generation 1698. mean fitness: 0.32. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634}xxxx.p..p.l.e..p..l..l..l.....l....e.l..e............l..e.l......l....
generation 1699. mean fitness: 0.34. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.043650793650793634,0.043650793650793634,0.1354497354497355}xxxxx.l.l..g....l.e...g..p.g..g...l...............p..l....l....l.p..g.p.e
generation 1700. mean fitness: 0.31. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.043650793650793634,0.1354497354497355,0.15714285714285714}xxxx..l..e..l.p..............g....l...l..p.l.....p.g...p.l....g.....e
generation 1701. mean fitness: 0.35. Best fitnesses: {0.043650793650793634,0.043650793650793634,0.15714285714285714,0.15714285714285714,0.16084656084656085}xxxxx..l..l..l.....l.l...........p....l..e.l....l.l.p.l......p...p..l..
generation 1702. mean fitness: 0.35. Best fitnesses: {0.043650793650793634,0.15714285714285714,0.15714285714285714,0.15714285714285714,0.16084656084656085}xxxx.p...e...p.lg..e......l..p.e.l.p......p..l.l.l....e......l.e.p.......
generation 1703. mean fitness: 0.36. Best fitnesses: {0.15714285714285714,0.15714285714285714,0.15714285714285714,0.16084656084656085,0.16084656084656085}xxxx..e....p....p.l.e.p..............l.....e....g.p.....e.lg...l.l..l.

[Treblesteph]

This bug would be much easier to detect if the fitness plot were working (i.e. if I can fix #35).

[Treblesteph]

Lots of commits in speedup branch, but I think this is a fairly recent problem so will start with the most recent. Also some commits have been neatening up code and adding comments so will not focus on these until I have first examined those which were meant to change the behaviour of the code.

I think the following areas are most likely to be the cause of the bug:

• commit 1ae67cf
  • jumping to next time point where something is likely to happen (according to multiples of lags) rather than simulating every single minute of time
• commit 8b65f1b
  • only running fitness calculation for simulations where the concentration timeseries is likely to have changed (using a mutated field in the Network type)
• commit 3acdf66
  • not rerunning simulations after crossovers between identical parents
• various commits (I was not being good at git here 👎)
  • only rerunning simulations after mutations which caused 'mutated' switch to be turned on