Fitness stagnates
6 replies
Fluke
8 years ago #114037
Hi, i have a question, i’m trying many settings, but I can not not to stagnate the fitness? What do you suggest to change to get to higher levels? I did not reach more than 0.70 of fitness, i attach the graph of fitness.
please do not suggest we do restart if the fitness stagnate, this is only start the process but I’m not exceed the limits, I want a serious advice.
geektrader
8 years ago #131918
Your fitness must not necessarily reach “1” to be perfect, it depends on your weights and which parameters you use for your custom fitness. Mine for example has it´s highest value at 0.70, it won´t go any higher than that as it depends on how I balance (weight) the various options into the custom fitness. So you don´t have to worry that you can´t reach “1”, it´s absolutely normal.
Fluke
8 years ago #131927
I did not ask this, i asked how to overcome this limitation.
geektrader
8 years ago #131928
Why would you want to overcome this “limitation”? It´s not limitation at all, your weighted fitness just might have a max of 0.7 or 0.8, just like mine, because you´ve composed it this way. SQ doesn´t normalize the fitness to 1 in any case, so your 0.7 = 1.0 if it would be normalized. You don´t need to “overcome” this, there is no benefit in this for you.
Fluke
8 years ago #131930
then what is the point to set a weighted fitness? I can leave it at random so who cares, as long as the equity is good, wrong!!! I want the strategies with my fitness correct, 0,6 of fitness does not make sense, I do not know what has followed my rule and what does not.
geektrader
8 years ago #131931
Sorry, I can´t explain it further to you it seems. Fact is, it is normal that if weighted, the fitness does not need to reach “1”. I have a function that just reaches 0.16 as it´s best value, and still it´s absolutely right and systems it finds with that goal are exactly what I wanted and specified in the weighted fitness function. It´s just that the fitness is not normalized to “1” when using the weighted fitness function. Maybe Mark can explain it better. In any case, it does not give any problem, if it stagnates for you at 0.7, then this is your “1” in this case and you can see the 0.7 as the best value that can be reached by your current fitness weights, which is absolutely normal and supposed to work like that.
seaton
8 years ago #132043
Some thing you could try is change randomness of mutation, Sometimes GAs can get “stuck” in a solution where its not necessity optimal , the Mutation may help overcome this to some extent and helps to “add new blood” to the gene pool to get the fitness out of a local maxima
Viewing 6 replies - 1 through 6 (of 6 total)