1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
|
# Copyright (C) 2016 EDF
# All Rights Reserved
# This code is published under the GNU Lesser General Public License (GNU LGPL)
import numpy as np
import StOptGeners
import StOptGlobal
import imp
# Simulate the optimal strategy , threaded version
# p_grid grid used for deterministic state (stocks for example)
# p_optimize optimizer defining the optimization between two time steps
# p_funcFinalValue function defining the final value
# p_pointStock initial point stock
# p_initialRegime regime at initial date
# p_fileToDump name of the file used to dump continuation values in optimization
def SimulateRegressionControlDist(p_grid, p_optimize, p_funcFinalValue, p_pointStock, p_initialRegime, p_fileToDump, p_bOneFile) :
try:
imp.find_module('mpi4py')
found = True
except:
print("Not parallel module found ")
found = False
if found :
from mpi4py import MPI
# from the optimizer get back the simulation
world = MPI.COMM_WORLD
simulator = p_optimize.getSimulator()
nbStep = simulator.getNbStep()
states = []
particle0 = simulator.getParticles()[:,0]
for i in range(simulator.getNbSimul()) :
states.append(StOptGlobal.StateWithStocks(p_initialRegime, p_pointStock, particle0))
toDump = p_fileToDump
if p_bOneFile is False:
toDump += "_" + str(world.rank)
ar = StOptGeners.BinaryFileArchive(toDump, "r")
# name for continuation object in archive
nameAr = "Continuation"
# cost function
costFunction = np.zeros((p_optimize.getSimuFuncSize(), simulator.getNbSimul()))
# iterate on time steps
for istep in range(nbStep) :
NewState = StOptGlobal.SimulateStepRegressionControlDist(ar, istep, nameAr, p_grid, p_grid, p_optimize, p_bOneFile).oneStep(states, costFunction)
# different from C++
states = NewState[0]
costFunction = NewState[1]
# new stochastic state
particles = simulator.stepForwardAndGetParticles()
for i in range(simulator.getNbSimul()) :
states[i].setStochasticRealization(particles[:,i])
# final : accept to exercise if not already done entirely
for i in range(simulator.getNbSimul()) :
costFunction[0,i] += p_funcFinalValue.set(states[i].getRegime(), states[i].getPtStock(), states[i].getStochasticRealization()) * simulator.getActu()
# average gain/cost
return costFunction.mean()
|
