Optimization in Strategy Tester
Hi,
I'm trying to rewrite my expert advisor from MT4 to MT5 but I have some problem with optimization.
The problem is that EA is based on neural networks and needs training. In MT4 I run my EA in optimization mode (e.g. variable "epochs", from 1 to 1000, step 1) to train neural networks. On every step my EA is saving current weights into binary file and that's why every new step depends on previous one.
In MT5 optimization Strategy Tester is using more then one core and is trying to do parallel computation which is useless for me.
Any idea how to solve the problem?
regards
The simplest is to disable the other cores, so that the Strategy tester only runs in one core. To disable the other cores: Strategy Tester>Agents>(right click on core to disable)Edit>(uncheck)Enable. Another approach is to create a unique weights file on each pass by mangling the filename with a datetime timestamp (works only if each test takes more than one second). Then you can combine weights files of previous timeframes, if desired. If you are using standard backprop for your ANN, then bulk (or combined error) update is certainly possible. Moreover, the bulk adjustment method was in fact the original design of backprop.
- Patrick
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hi,
I'm trying to rewrite my expert advisor from MT4 to MT5 but I have some problem with optimization.
The problem is that EA is based on neural networks and needs training. In MT4 I run my EA in optimization mode (e.g. variable "epochs", from 1 to 1000, step 1) to train neural networks. On every step my EA is saving current weights into binary file and that's why every new step depends on previous one.
In MT5 optimization Strategy Tester is using more then one core and is trying to do parallel computation which is useless for me.
Any idea how to solve the problem?
regards