3

I was reading Stanley's paper but I couldn't figure out what exactly are Disjoint and Excess genes in NEAT. I understand they appear to be related in some particular way with the fact that all of them contain innovation numbers not pertaining to both parents. But what distinguishes them?

Could anyone shed some light into the issue?

devoured elysium
  • 101,373
  • 131
  • 340
  • 557

1 Answers1

5

When aligning two genomes by gene ID (innovation number), the mismatches at the ends are referred to as excess genes, and all other mismatches are referred to as disjoint genes. As far as I am aware no NEAT implementation has ever treated disjoint and excess genes differently. The distinction was made in early NEAT papers, most probably because treating the types of mismatches differently was being suggested as a possible future research topic.

(FYI - speaking as the author of SharpNEAT).

redcalx
  • 8,177
  • 4
  • 56
  • 105
  • I was having the same question and couldn't find a descent answer nor could i think of any reason to make the distinction in calculating the distance between 2 chromosomes / genomes. Do you know if any additional research was done meanwhile into the effect of adjusting the coefficients of excess vs disjoint genes? – Maurice CGP Peters Feb 01 '18 at 13:13
  • multineat implementation treats them differently: they have different coeffs in compatibility distance function – AlexStepanov Nov 30 '18 at 08:18
  • If there is no good reason for it, it's an arbitrary choice. I implemented NEAT from scratch and everything seems to work fine when I treat any genes without the same innovation number as disjoint. Instead of `c1*excessGenes/totalGenes + c2*disjointGenes/totalGenes + c3*averageWeightDifferences` for genetic distance, I use `c1*disjointGenes/totalGenes + c2*averageWeightDifferences` Using the classic example from here ( https://www.cs.cmu.edu/afs/cs/project/jair/pub/volume21/stanley04a-html/node3.html ) there'd be 5 disjoint genes. PS: don't make the mistake of assuming paper authors are gods. – Austin Capobianco Mar 31 '20 at 21:20
  • 2
    "PS: don't make the mistake of assuming paper authors are gods." This is worth emphasising - the excess/disjoint distinction was an idea that was never explored, and there's no reason to believe it has any merit. – redcalx Apr 06 '20 at 10:20
  • 1
    Also, it seems like there is no structural significance of the distinction between disjoint and excess genes. If two connections connect the same nodes, they will be matched and if not they can't be matched, and being disjoint or excess doesn't give any additional information, IMO. Moreover, the same gene can be an excess or disjoint gene depending on the time it emerged, which isn't informative. I think this should be simplified to Matched and Unmatched genes, unless there is something I overlook. – Acemad Jun 20 '21 at 08:55