-
Notifications
You must be signed in to change notification settings - Fork 103
Benchmarks
At strategy_parameters
folders there are by default some parameter for GA benchmark,
so we can run them:
$python3 japonicus.py -egc --strat griewangk
# this runs benchmark mode on griewangk function
# note the -e flag.
Such functions have lowest point as Zero, and japonicus should grab its settings
and find the parameters that deliver profit = 0
.
This is useful to test new GA parameters, representation and generation processing flows. Works as a check to see if the ensemble will get stuck on local maximas or will look far enough to be able to find new paths to the best set of parameters.
Benchmark mode is of little use for the average user. It's more for people who like genetic algorithm development and are willing to test new settings and push commits to the GA system of japonicus.
Currently, the implemented GA system is not perfect but seems good enough for the job on trading bots.