Why Auto-Evo is Dead

~sciocont wrote:

roadkillguy wrote:

How would you suggest to implement this in the current outline?

I'm not exactly following what you mean.

Basically, I'm saying that a species can only compete for one niche other than its own at a time.

QFT

For simplicity's sake, yes. The population routine will take care of the "I'm only competing for one little resource" bits of competition, affecting our populations without us messing about with running multiple competitions.

I'm confused about how the competing and resources tie in.

What do you mean by those?

Two species both share the primary food source?

roadkillguy wrote:

I'm confused about how the competing and resources tie in.

What do you mean by those?

Two species both share the primary food source?

Yeah, basically. Competition is just two species needing the same thing to survive.

What's the problem then?

Implementation? I'm not really sure where this conversation has gone, It's getting hard for me to follow.

Quote :

Species compete for resources.

How would we implement that?

Because when we're talking "primary food source" we're oversimplifying. Most species can eat a wide variety of items, and have other necessary resources such as places to make dens/nests, as well as some mineral requirements.

Basic implementation would be that each species has a required resource density: how much available food per square unit, access to minerals, and acess to shelter/nesting sites, as well as a tag for how much water it needs. The amount of resources determines the population, and the population determines how many resources will be available at the next check.

Primary food source is for use with calculateNewPopulation() and calculateFitness().

See, we need to be able to make things moddable. If the modders don't have access to built in variables, they cant mod.

What I'm asking is, how would those variables you specified be stored and accessed by the niche mods?

roadkillguy wrote:

Primary food source is for use with calculateNewPopulation() and calculateFitness().

See, we need to be able to make things moddable. If the modders don't have access to built in variables, they cant mod.

What I'm asking is, how would those variables you specified be stored and accessed by the niche mods?

I'm simply assuming that tagged variables will be stored in a table somewhere where they can be viewed/modded/copied... I'm an Enviromental Scientist, not a programmer. I'm learning a tiny bit of modding, but I can't program in the least.
Primary food source should be more like an average of a couple different food sources, no more than 2 or three, but other than that, okay.

We could have a top food sources list with maybe 5 food sources ordered from greatest to least, that would work for sure.

The only problem with this is, how do we calculate the new population with 5 food sources? If each trophic level only gets 10% or so of the trophic level below it, how do we know which population/trophic level to look at for population? This method leaves this problem up to the modders.


Things like competition would have to be run within the simulation, and IMO shouldn't be included within auto-evo. The simulation may have AI flaws, and they shouldn't be recorded and ruin the auto-evo system.

roadkillguy wrote:

We could have a top food sources list with maybe 5 food sources ordered from greatest to least, that would work for sure.

The only problem with this is, how do we calculate the new population with 5 food sources? If each trophic level only gets 10% or so of the trophic level below it, how do we know which population/trophic level to look at for population? This method leaves this problem up to the modders.


Things like competition would have to be run within the simulation, and IMO shouldn't be included within auto-evo. The simulation may have AI flaws, and they shouldn't be recorded and ruin the auto-evo system.

Yeah, I agree that we should keep auto evo as simple as possible.

bashi this wil maybe help you:
http://evolutionandgames.nl/wiki/index.php?title=Main_Page
i hope you wil find something useful, but i aren't even sure is here anything helpful.
so just check is here, it could be.

dimasciene wrote:

bashi this wil maybe help you:
http://evolutionandgames.nl/wiki/index.php?title=Main_Page
i hope you wil find something useful, but i aren't even sure is here anything helpful.
so just check is here, it could be.

This looks pretty useful. By which I mean it looks relevant and very, very complicated, which I assume means it's useful.

dimasciene wrote:

bashi this wil maybe help you:
http://evolutionandgames.nl/wiki/index.php?title=Main_Page
i hope you wil find something useful, but i aren't even sure is here anything helpful.
so just check is here, it could be.

In case you haven't noticed, Bashi does not participate in the forums anymore.

Alright, it's been a while but I think I should clarify some things about what Seregon and I are working on with auto-evo.

The auto-evo is going to be set up primarily as a population dynamics model.
I.E. for a species X with population P_x and food source Y with amount in system P_y a very basic equation would be.

dP_x/dt = a*X*P_y-b*X

where dP_x/dt is the change in the population of species X with respect to time. a is a paramter linking food interactions with X and how much growth X gets from consuming Y. b is a death coefficient which relates to how fast each member of a species dies off (lifetime).

Now this is a very simple model, and you can imagine a more complex one would have interactions between every species and food source, and some things would be dependent on time (such as sunlight available during a year cycle). This makes the problem a very large matrix, but it is literally just a huge plug and chug problem for the computer, which is what computers are good at.

Now what does this have to do with the auto-evo?

The auto-evo seeks to adjust the paramaters a,b, etc. to maximize the growth dP_x/dt for the species that have been selected to "evolve". This is done using partial differentiation techniques which Sergeon is much better at than I am, but it does work, and is also fairly simple to use.

The paramaters are the key point in this. They relate to the properties of each species, and must be balanced so that increasing one would have a negative effect on others. In a sense, this will make "optimum values' for interacting which the species will slowly move towards, and these optimum's will change depending on environment.

Where we are at now on auto-evo.

We are currently working on how to balance the parameter interactions to get realistic results, and deciding which properties will affect which parameters.

Here is the post we were discussing this on if you want more information, or just want to contribute more ideas. Population Dynamics,

So I guess the first step would be agreeing on a set of properties for each species. I.E. Food consumption, toxins, toxin resistance, aggressiveness, mass, lifespan, speed, camoflouge, etc.

The next step would be to develop equations to relate these to the parameters in the population dynamics model. (Much easier said, than done). But once that is done auto-evo is pretty complete, with maybe a few tweaks to improve its realism.

So all in all, I'd say auto-evo is far from dead.

Nice work gdt. This all lines up very nicely with what I've been developing, and I think it can all be combined (with a little work) into the current auto evo concept. I haven't got time this week to have a proper think about this, or develop it any further, but I'll try to do so over the weekend, or early next week.

I also wanted to point out that this thread is largely out of date, the current auto-evo concept is here. It's rather a long read, but worthwhile to see how we got where we are now.

There's an easy way to resolve such a problem. You create a base chance value (that will undoubtedly get tweaked) and create a mathematical equation to automatically apply environmental variables to it. For example, if planetary temperatures reach extremes (ice age, etc.) then it lowers the chance that a creature automatically evolves. It's not a matter of recreating evolution, our goal is to SIMULATE the effect as best as we can.

TropicalMammoth wrote:

There's an easy way to resolve such a problem. You create a base chance value (that will undoubtedly get tweaked) and create a mathematical equation to automatically apply environmental variables to it. For example, if planetary temperatures reach extremes (ice age, etc.) then it lowers the chance that a creature automatically evolves. It's not a matter of recreating evolution, our goal is to SIMULATE the effect as best as we can.

Tropical this has already been discussed. You seem to have a habit of necro bumping.

Its fine, this was only a couple months ago and is a very relevant and significant topic. Whats important is that he, or any other new people, come up to date with the current concept, but I'm not one for explaining Auto-Evo, so I'll let one of the others do that.