mango.optimize.distributed_minimize

mango.optimize.distributed_minimize(fun, *args, **kwargs)

Wrapper for the scipy.optimize.minimize() function for distributed function evaluation on multiple MPI processors. The fun argument should be an instance of DistributedMetricEvaluator. This function should be called from all MPI processes.

Parameters:

• fun (DistributedMetricEvaluator) – Function which is to be minimized.
• bcastres (bool) – If True (default is True) the result object (scipy.optimize.OptimizeResult) returned by scipy.optimize.minimize() is broadcast from the fun.root process to all other fun.comm MPI processes. If False only the fun.root process returns the result and other processes return the None value.

Example:

import mango.mpi
import scipy as sp
import mango.optimize

def my_func(x):
   return (mango.mpi.world.Get_rank()+1) * (x * x + x + 1)

# dfunc sums/reduces (rank+1)*(x*x + x + 1) from all processes to the dfunc.root process
dfunc = mango.optimize.DistributedMetricEvaluator(my_func) 
x0 = 8
res = mango.optimize.minimize(dfunc, x0, method="Powell")
print("res.x = %s, res.fun = %s" % (res.x, res.fun))