| trade.optimizer.control {PortfolioProbe} | R Documentation |
Sets parameters that control the optimization algorithm in
trade.optimizer (and hence in random.portfolio.utility).
trade.optimizer.control(iterations.max = 20, fail.iter = 0,
funeval.max = .Machine$integer.max, trace = TRUE,
exit.obj = -big, doubleconst = FALSE, runs.init = 3,
runs.final = 2, runs.min = 1, stringency = 0,
nonconverge.mult = 2, feasible = 0, miniter = 0,
force.risk.aver = FALSE, enforce.max.weight = TRUE,
save.iterhistory = FALSE, ...)
iterations.max |
integer giving the maximum number of iterations to perform in a run. |
fail.iter |
integer giving the maximum number of consecutive iterations that
fail to improve the solution without the algorithm stopping and
declaring that convergence has been achieved.
For example, if fail.iter is 5, then the algorithm continues
with 5 consecutive failures but stops upon the 6th consecutive
failure.
|
funeval.max |
integer giving the maximum number of function (portfolio) evaluations to perform. The most common use of this is to set it to one or zero when you have a specific trade that you would like to have done. |
trace |
logical value.
If TRUE, then information on the progress of the optimization
is printed.
This is ignored under Windows – tracing information is not
available on this platform except in BATCH mode.
|
exit.obj |
number such that the optimization terminates if the best solution has
an objective at least as good (i.e., an objective smaller than
or equal to exit.obj).
(Useful for comparing algorithms,
but otherwise probably uninteresting.)
|
doubleconst |
logical value.
If TRUE, then any asset which goes from long to short or
vice versa in a trade is subject to both intercepts of the trading
costs.
If FALSE, then the first intercept is used
but the second is not.
This is ignored if cost.par is given (or there are no costs).
|
runs.init |
if stringency is greater than zero, then the maximum number
of runs to do on the original problem.
|
runs.final |
if stringency is greater than zero, then the maximum number
of runs to do with the trading universe restricted to those assets
that traded in at least one of the initial runs.
|
runs.min |
if stringency is greater than zero, then the minimum number
of runs that are to be performed.
|
stringency |
an integer giving one less than the number of runs that need
to match as the best solution before the optimization is exited.
If stringency is 0, then only one run is done.
If stringency is 1, then the best solution needs to occur
twice before the optimization stops.
If the best solution does not occur enough times before the maximum
number of runs is exhausted, then a final non-convergence run is
performed that starts from the best solution found.
|
nonconverge.mult |
if stringency is greater than zero and the stringency
requirement is not met, then a final run is performed starting
with the best solution found.
nonconverge.mult gives the ratio of iterations allowed for
this final run relative to the iterations allowed in the others.
So if iterations.max is 5 and nonconverge.mult is 3,
then the non-convergence run is allowed up to 15 iterations.
|
feasible |
if all constraints are not met at the end of the iteration equal to
the value of feasible, then the optimization is aborted.
But if feasible is 0, then the optimization does not
stop on account of broken constraints.
|
miniter |
the minimum number of iterations to be performed in a run even
if fail.iter says to stop already.
|
force.risk.aver |
logical value which is only used when the utable
argument is explicitly given and optimization with risk aversion
is performed.
If TRUE, then the risk.aversion argument overrides
the risk aversion values in the utility table.
|
enforce.max.weight |
logical value; if TRUE, then forced trades are automatically
created if any positions in the existing portfolio break their
maximum weight constraint.
This doesn't absolutely guarantee that the constraints will be
met, but they generally will be unless the gross value is given a
lot of latitude.
|
save.iterhistory |
logical value; if TRUE, then a vector of the objective value
at each iteration of the optimization will be returned.
|
... |
there are additional arguments that make this function compatible with
random.portfolio.control.
|
a list with the following components:
icontrol |
vector of the integer-valued control parameters. |
dcontrol |
vector of the double precision control parameters. |
aux |
vector of the auxiliary control parameters. |
This help was last revised 2010 January 02.
trade.optimizer,
random.portfolio.utility,
random.portfolio.control.
my.to.control <- trade.optimizer.control(iterations=100, fail=5,
stringency=1, runs.init=5, runs.final=3)
trade.optimizer(eq.prices, varian, control=my.to.control)