{{{173876511841242643565349714506270612236,4,1},107},{{201412842028162214137229450141139961424,4,1},136}}
In[]:=
GraphicsRow[Table[PlotCA[PerturbCA[{getca[{173876511841242643565349714506270612236,4,1}],{173876511841242643565349714506270612236,4,1}},"ReturnPerturbations"->True]],20]//QuietEcho]
Out[]=
In[]:=
With[{ru={173876511841242643565349714506270612236,4,1}},PerturbationSensitivityPlot[{getca[ru],ru}]]
»
False
Out[]=
In[]:=
With[{ru={201412842028162214137229450141139961424,4,1}},GraphicsRow[Table[PlotCA[PerturbCA[{getca[ru],ru},"ReturnPerturbations"->True]],20]//QuietEcho]]
Out[]=
In[]:=
With[{ru={201412842028162214137229450141139961424,4,1}},PerturbationSensitivityPlot[{getca[ru],ru}]]
»
False
Out[]=
Multiple perturbations
Multiple perturbations
Bigger Picture
Bigger Picture
You can’t try every perturbation in the adaptive evolution ... so in a sense the system has to “generalize” ; i.e. we trained it on certain perturbations, but others will show up in its actual life.
If you try to insist on being robust to every perturbation ...... you end with trivial evolution (?)
Let’s say the robustness probability in training is p ... presumably that’s also the robustness when running
[cf some organisms spread a gazillion eggs etc , expecting only a few will work out]
[cf some organisms spread a gazillion eggs etc , expecting only a few will work out]
So when things go bad after a perturbation, we’re being knocked out of an attractor
Things to Look At
Things to Look At
Genetic changes
Genetic changes
Can we change something in the rule and not have everything fail?
Global changes
Global changes
Algorithmic drugs
Algorithmic drugs
All possible changes
All possible changes
How many distinct developed patterns are there?