The genetic algorithm and its possible applications - page 15
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
"If you stare into the abyss for too long, the abyss begins to stare into you.
If you suffer long enough, something will come out).
138 messages with no application to forex trading. Optimizer in the form that works now simulates the curve of the price chart (not for nothing I studied at the Mathematical Modeling Department, we used such methods for mathematical description of technological processes to predict with an acceptable error the behavior of these processes in the near future, when we get a new real result (almost the same as predicted) we add it to the previous optimization to get a new result in the future) - such models are applicable to the near future and to the future. (by the way such methods are used to build missile trajectories with obstacle avoiding and threat avoiding - there just use mathematical modelling methods to get a mathematical model from the beginning, then use artificial neural network to model mathematical model based on Ins - this is needed because Ins calculations are hundreds times faster than original mat model, what affects time of missile reaction to obstacles and threats + resulting model is put on physical level in the The genetic algorithm itself isn't bad, but I'm trying to use it not to model the price curve, but to model the behaviour of the price chart at least in certain timeframes and in certain exchange conditions. These constraints plus the price chart itself is the outside world for the resulting model - the model itself is constrained by that world. That is, the model is always adapted to the conditions of the outside world, which we can describe due to our knowledge. By the way, humans are also fitted to the environment - we cannot be in outer space, for example - for those who are engaged in massive philosophical arguments about order and chaos. Artificial neural network is certainly good, if you have already made attempts to model a strategy for exchange with its help (and such models exist, and cost a lot of money, I read them when writing about artificial neural networks). I would like to read something useful for practical use in trading. All this means that - try to apply the genetic algorithm in the form offered in this forum, I have obtained modest results, but I will not share them. My models are limited by my lesser knowledge of trading compared to yours as I've been trading for less than a year. I would like more knowledge from you!
It is very difficult to read your text, please divide it into sentences and paragraphs.
It is very difficult to read your text, please divide it into sentences and paragraphs.
It all makes sense.
There is a problem with frames on 'big' genetics.
We will fix it.
With 96-bit genetics, there's still the problem of handing out tasks. After a normal first generation pass, half of the local agents remain finised until the end. Nothing in the logs. When reducing the number of variants to 64bit everything is ok.
Try it with any EA. If it doesn't reproduce, I'll pass.
With 96-bit genetics, the problem of handing out tasks remains. After a normal first generation pass, half of the local agents remain finised until the end. Nothing in logs. When reducing number of variants to 64bit everything is ok.
Try it with any EA. If it doesn't reproduce, I'll pass.
Turn on the checkbox of full logs in the context menu of the tester's log. There should be a record of how many which jobs go where.
I am attaching the log from b2368. I don't see anything there about assigning tasks.
Checked in newly appeared b2374, also does not work. For now, I'm increasing optimization step to 64 bits. When I find an opportunity, I reduce the number of optimized variables. I understand that a large number of sets is wrong as there is nothing but local nodes to be found. I just need to estimate optimization ranges and interrelation of variables.
I am attaching the log from b2368. I don't see anything there about assigning tasks.
Checked in the newly appeared b2374, also does not work. For now, I'm increasing optimization step to 64 bits. When I find an opportunity, I reduce the number of optimized variables. I understand that a large number of sets is wrong as there is nothing but local nodes to be found. I just need to evaluate the optimization ranges and interrelation of variables.
Thank you for the logs. The problem is clearly visible. Let's solve it
You are saying that there is no such a problem on 64-bit genetics and all cores are loaded evenly?
Are you saying that on 64-bit genetics this problem does not exist and all cores are loaded evenly?
Absolutely. Well, apart from the fact that by the end of the generation everyone is done, and one agent alone does up to a few dozen more passes. There's no redistribution of the job package, but that's a general problem, not just 96-bit. I understand that giving one agent a single job would be inefficient for cloud agents, but maybe for local agents adopt a different strategy - either give a minimum of jobs, or dynamically redistribute when idle (for example, if more than a quarter of agents are idle and the busiest agent has more than N jobs left).
Absolutely. Well, apart from the fact that by the end of the generation everyone is done, and one agent alone makes up to a few dozen more passes. There's no job packet redistribution, but that's a common problem, not just 96-bit. I understand that giving one agent a single job would be inefficient for cloud agents, but maybe for local agents adopt a different strategy - either give a minimum of jobs, or dynamically redistribute when idle (for example, if more than a quarter of agents are idle and the busiest agent has more than N jobs left).