Objective.ox

Objective is the base class for defining what to optimize or solve.

The base and derived objective classes are documented here. Click on "Hierarchy" on the menu bar to see them organized in a logical order.

Global enumerations

Enumerations
MixedWeightOptions		Tags weighting options.
QuasiAlgorithms		Tags for Gradient-based optimization algorithms.

BlackBox : UnConstrained : Objective

Represents a blacbox objective.

This is the container class for a standard function to maximize.

A BlackBox has no internal structure (such as separability of parameters).

Public methods
BlackBox Create a blackbox objective.

Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

CES : BlackBox : UnConstrained : Objective

Constant elasticity of subsitution function.

Public fields
A
alphas
elast
x
xpon
Public methods
CES		Create a constant elasticity of subsitution objective.
vfunc	virtual	Compute CES.

Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

CobbDouglas : BlackBox : UnConstrained : Objective

Cobb-Douglas objective. $$f = A \prod_{i=0}^{N^-} x_i ^{\alpha_i}.$$

$A$ and $\alpha$ are CV-compatible objects.

Public fields
A

alphas

x

Public methods
AnalyticGradient virtual

CobbDouglas Create a CobbDouglas objective.

vfunc virtual $A{\prod}_{i=1}^N x_i^{\alpha_i}$.

Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

Constrained : Objective

Container for Constrained Objectives.

Public methods
CheckPoint	virtual	Store current state of a Constrained Objective (checkpoint to disk).
Constrained		Create a constrained objective.
equality	virtual	Default Equality Constraints.
funclist		Compute the objective and constraints at multiple points.
Gradient	virtual	Compute the ∇f(), objective's gradient at the current vector.
inequality	virtual	Default InEquality Constraints.
Jacobian	virtual	Compute the Jg(), vector version of the Lagrangian's Jacobian at the current vector.
Lagrangian		Decode the input, return the whole vector, inequality and equality constraints, if any.
Merit	virtual

Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, contour, Decode, Encode, fobj, FreeStatus, Hessian, Interact, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

CPoint

Public methods
Vec

DataObjective : BlackBox : UnConstrained : Objective

Base class for automatically generated econometric objectives: likelihood functions and GMM objective.

This objective takes a "data" object that is typically either a OutcomeDataSet or a PredictionDataSet.

Public fields
data const path or prediction object.

stage estimation stage.

tplist transition parameters .

uplist utility parameters .

Public methods
DataObjective An objective based on an economic model with data and (possibly) a nested solution method.

SetStage Set the estimation stage.

TwoStage Specify the objective as two stage.

vfunc virtual Calls and returns data->EconometricObjective() or uses SemiClosedForm() if in Gradient calculation.

Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

Equilibrium : System : Objective

A nonlinear system for computing equilibrium. This provides a framework for equilibrium prices in a model with an aggregate production function.

$T$ periods (default $T=1$)

Production

A $N\times 1$ vector $X.$

$F(X^d)$ is aggregate output where $X^s$ stands for aggregate demand.

$p_t$ is a $n\times 1$ price vector in period $t$.

The respresentative firm acts to maximize static profit by choosing $X^d_$ accounting for input prices and a depreciation vector $\Delta$. First order conditions at $t$ are: $$\nabla F(X^d_t)^\prime - \Delta - p_t = \overrightarrow{0}$$ where $\nabla F$ is the $1\times N$ gradient of the production function with respect to $X^d_t$.

Households

Consumer behavior is described by a DP with $T$ periods. If the clock is ergodic then it is a stationary equilibrium for infinitely lived agents. If the clock is non-stationary then it is a sequence of $T$ static equilibrium problems which are connected because agent behavior at $t$ depends on future prices and past choices through the current distribution across states. (This version does not handle overlapping generations.)

Adding variables to $\gamma_r$ (random effects) allows for heterogeneity. (This version does not allow fixed effects.)

The price matrix $p$ enters $U()$, usually through a budget constraint that determines consumption.

Quantities of per-capita values supplied by households ($X^s$) are determined by the predicted amounts : $$X^s_t = \sum_{\gamma_r}g(r)\left[ \sum_{\theta\in\Theta} f_t(\theta;\gamma_r) x_t(\theta;\gamma_r)\right].$$ Here $x_t(\theta;\gamma_r)$ is the vector of elements of of the DP outcome $Y$ that match up to the factors of production. $f_t$ is the distribution over states at $t.$ The user specifies the intial distribution $f_0$. If the equilibrium is stationary then this is the ergodic distribution $f_\infty$.

A fixed factor can be included by adding a constant AuxiliaryOutcome to to the DP, which would then always average across states and household types to the fixed factor.

In other words, $X^s_t$ are computed from the PathPrediction of length $T$ starting from exogenous initial conditions.

Equilibrium

In equilibrium the price vector $p^\star_t$ satisfies the first order conditions when $X^s$ is inserted for $X^d$ at each $t.$

In general, the full price vector $p^\star$ is the solution to a system of $TN$ equations. Depending on the production fucntion $f(X)$ some of the equations may be solved out to reduce the number of equations. In that case some elements of $p^\star$ are Determined parameters not available for the system algorithm to vary.

Public fields
aggF	const	list of aggregate prod.
deprec	const	0.0 or vector of quantity depreciations.
foc		computed system.
Q		aggregate Qs from prediction.
Qcols	const	columns of pred that match up to inputs.
stnpred	const	ergodic distribution PathPrediction.
T	const	number of periods in equilibrium.
Public methods
Equilibrium		Create a system of aggregate equilibrium equations.
Print		Prints a message and details about the objective.
vfunc	virtual	Built-in system of equations for Equilibrium models.

Inherited methods from System:: EndOneDim, equations, SetOneDim, System
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Gradient, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from System:: eq1, normexp
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

GHKMNP : MNP : MultiNomialChoice : BlackBox : UnConstrained : Objective

Correlated MNP using GHK .

Public fields
ghk	const
sigfree	const
SigLT	const
Public methods
GHKMNP		GHK based objective for MNP log likelihood.
SetGHK
vfunc		Compute and return the vector of log-likelihoods at the current parameters.
Enumerations
Anonymous enum 1

Inherited methods from MultiNomialChoice:: Estimate, MultiNomialChoice, SetD
Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from MultiNomialChoice:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

GQMNP : MNP : MultiNomialChoice : BlackBox : UnConstrained : Objective

Independent MNP using Gausss-Hermite integration.

Public fields
Npts	const	# of nodes in quadrature
Public methods
GQMNP		Gauss-Hermite based objective for MNP log likelihood.
vfunc		Compute and return the vector of log-likelihoods at the current parameters.

Inherited methods from MultiNomialChoice:: Estimate, MultiNomialChoice, SetD
Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from MultiNomialChoice:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

MixPoint

Public methods
aggregate

Mixture : Separable : UnConstrained : Objective

Public fields
D	const	# observed types, environment
DK	const	D x K
dkfree
DKL	const
Dvar	const	Discrete environment-type var
FinL
Flabels
Included		matrix of indicators of valid d,k combos
lfree
mcur	const
nfree
WStart
Public methods
CheckMax	virtual	If `scur.v > maxpt.v` call Save and update `maxpt`.
Decode	virtual	Decode and return the parameter array.
Encode	virtual	.
fobj		Decode the input, compute the objective, check the maximum.
funclist	virtual	Compute the objective at multiple points.
Gradient	virtual	Compute the ∇f(), objective's gradient at the current vector.
IncludedDK		Indicate with (d,k) combinations should be computed.
Jacobian	virtual	Compute the Jg(), vector version of the objective's Jacobian at the current vector.
Mixture		Create a mixture objective.
vfunc	virtual
vobj	virtual	Decode the input, return the whole vector.
WDecode
WEncode
Wfunclist	virtual

Inherited methods from Separable:: Common, Separable
Inherited methods from UnConstrained:: UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMaxV, CheckPoint, contour, FreeStatus, Hessian, Interact, Load, Menu, Objective, Parameters, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams

Inherited fields from Separable:: C, CDNV, ComInd, K, kfree, KL, kNvf, Kvar, NvfuncTerms, scur
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nstruct, once, p2p, PsiL, PsiType, RunSafe, subp2p, vcur, Volume, Warned

MLogit : MultiNomialChoice : BlackBox : UnConstrained : Objective

Multinomial Logit Model.

Public methods
MLogit		Create a multinomial logit model.
vfunc		Return the log-likelihood vector.

Inherited methods from MultiNomialChoice:: Estimate, MultiNomialChoice, SetD
Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from MultiNomialChoice:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

MNP : MultiNomialChoice : BlackBox : UnConstrained : Objective

Container for normally-distributed multinomial discrete choice reduced-form models.

Inherited methods from MultiNomialChoice:: Estimate, MultiNomialChoice, SetD
Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from MultiNomialChoice:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

MultiNomialChoice : BlackBox : UnConstrained : Objective

Objective for multinomial choice models.

Public fields
betas	const	Array of J-1 parameter blocks, one for each equation except Y=0.
D		matrix of XB indices.
indY	const	N x J matrix of permutations the first column is index of Y second column is first non-Y choice third column is second non-Y choice etc.
J	const	J : number of options
Jvals	const	integer codes of choices
lk		vector or matrix of likelihods.
namearray	const	Labels for variables
NN	const	indexing vector
nX	const	K : number of X variables
X	const	X : matrix of exog variables
Y	const	Y : discrete endog vector
Public methods
Estimate
MultiNomialChoice		Create a new Multinomial Choice Model model.
SetD		Update $XB$ matrix of indices by observation and option.

Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

NoObjective : BlackBox : UnConstrained : Objective

Public fields
model
modelmethod
v
Public methods
NoObjective		A wrapper that acts like an objective but just calls a model's Solve method and returns 1.0.
vfunc

Inherited methods from BlackBox:: BlackBox
Inherited methods from UnConstrained:: Gradient, UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

Objective

Base class for objective optimization and system solving.

Public fields
DoNotConstrain		This objective uses raw parameter values.
dStepLinearGap	static const
EXT	static const	Extension for Load and Save files, =optobj.
Flabels		array free parameter labels
fname	const	Name for Load & files
fshold		holds DoNotVary vector for temp.
INGRADIENT		TRUE when computing Gradient, can be used by semi-closed form gradient methods.
L	const	label.
logf		log file
lognm	const	name of log file
maxpt	const	Best so far
MyVersion	static	Version of niqlow that code is written for.
newmax		Set in CheckMax(), TRUE if latest check was new max.
nfree		length of F in current cycle.
nstruct		length of X.
NvfuncTerms		Longest vector returned by vfunc(), default=1.
once		TRUE if encoded once
p2p		P2P object for using MPI across param vectors
PsiL		all parameter labels
PsiType		parameter class names
RunSafe		TRUE (default): exit if NaNs encountered during iteration FALSE: exit with `IterationFailed`
Start		Initial vector after Encode()
subp2p		P2P object for MPI on a single param vector.
vcur	const	current point.
Volume		Output Level.
Warned	static	.
Public methods
AggSubProbMat	virtual
Block	virtual
CheckMax	virtual	If `vcur.v > maxpt.v` call Save() and update `maxpt`.
CheckMaxV		Compare v to maxpt.
CheckPoint	virtual	Load or save state of the problem to a file.
contour	virtual	Graph the level curves of the objective in two parameters.
Decode	virtual	Decode a vector of free variables.
Encode	virtual	Encode vector of structural parameters.
fobj	virtual	Decode the input, compute the objective, check the maximum.
FreeStatus		Preserve the state of free or fixed status of the parameter vector.
funclist	virtual	Compute the objective at multiple points.
Gradient	virtual	Compute the ∇f(), objective's gradient at the current vector.
Hessian	virtual	Compute the Hf(), Hessian of objective at the current vector.
Interact	virtual
Jacobian	virtual	Compute the Jg(), vector version of the objective's Jacobian at the current vector.
Load		Load state of the problem.
Menu	virtual
Objective		Create a new objective.
Parameters	virtual	Add Parameters to the parameter list to be optimized over.
Print	virtual	Prints a message and details about the objective.
Recode	virtual	Reset the free parameter vector and the complete structural parameter vector.
ReInitialize	virtual	Revert all parameters to their hard-coded initial values.
ResetMax	virtual	Reset the value of the current and maximum objective.
Save		Store current state (checkpoint to disk).
SetAggregation		Set the formula .
SetVersion	static	Checks the version number you send with the current version of niqlow.
ToggleBlockElements	virtual	Toggle DoNotVary for one or more parameters.
ToggleParameterConstraint		Toggle the value of DoNotConstrain.
ToggleParams	virtual	Toggle DoNotVary for one or more parameters.
vfunc	virtual	Built in objective, f(ψ).
vobj	virtual	Decode the input, return the whole vector.

OneDimSystem : System : Objective

A One Dimensional Non-linear system (can be solved with OneDimSolve).

Public fields
ipar
isys
msys	const
Public methods
CheckMax		This is misnamed.
OneDimSystem		Create a new non-linear equation.
ResetMax		Reset the value of the system.
SetOneDim
vfunc	virtual

Inherited methods from System:: EndOneDim, equations, System
Inherited methods from Objective:: AggSubProbMat, Block, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Gradient, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vobj

Inherited fields from System:: eq1, normexp
Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

Separable : UnConstrained : Objective

Represent sum of K BlackBox objectives.

Public fields
C		Number of common parameters
CDNV
ComInd		Vector of indices into Psi of common pars
Flabels		.
Included
K	const	# unobserved types, sub-problems
kfree		K Vector of free specific parameters
KL	const	labels for types / sub problems
kNvf
Kvar	const	Discrete sub-problem var
nfree		Total free parameters
NvfuncTerms		Longest vector returned by vfunc(), default=1.
scur	const
Public methods
CheckMax	virtual	If `scur.v > maxpt.v` call Save and update `maxpt`.
Common	virtual	Add parameters to the objective that will have a single value across types/sub-problems.
Decode	virtual	Decode and return the structural parameter matrix.
Encode	virtual	Encode matrix of structural parameters.
fobj	virtual	Decode the input, compute the objective, check the maximum.
funclist	virtual	Compute the objective at multiple points.
Gradient	virtual	Compute the ∇f(), objective's gradient at the current vector.
Jacobian	virtual	Compute the Jg(), vector version of the objective's Jacobian at the current vector.
Print	virtual	.
Separable		Create a separable objective.
vfunc	virtual
vobj	virtual	Decode the input, return the whole vector.

Inherited methods from UnConstrained:: UnConstrained
Inherited methods from Objective:: AggSubProbMat, Block, CheckMaxV, CheckPoint, contour, FreeStatus, Hessian, Interact, Load, Menu, Objective, Parameters, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nstruct, once, p2p, PsiL, PsiType, RunSafe, subp2p, vcur, Volume, Warned

System : Objective

A non-linear system of equations to solve.

Public fields
eq1		1 equation of system.
normexp
Public methods
EndOneDim
equations	virtual	Default system of equations: vfunc().
SetOneDim	virtual
System		Create a new system of equations object.

Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Gradient, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

UnConstrained : Objective

Container for Unconstrained Objectives.

Public methods
Gradient	virtual	Compute the ∇f(), objective's gradient at the current vector.
UnConstrained		Create an unconstrained objective.

Inherited methods from Objective:: AggSubProbMat, Block, CheckMax, CheckMaxV, CheckPoint, contour, Decode, Encode, fobj, FreeStatus, funclist, Hessian, Interact, Jacobian, Load, Menu, Objective, Parameters, Print, Recode, ReInitialize, ResetMax, Save, SetAggregation, SetVersion, ToggleBlockElements, ToggleParameterConstraint, ToggleParams, vfunc, vobj

Inherited fields from Objective:: DoNotConstrain, dStepLinearGap, EXT, Flabels, fname, fshold, INGRADIENT, L, logf, lognm, maxpt, MyVersion, newmax, nfree, nstruct, NvfuncTerms, once, p2p, PsiL, PsiType, RunSafe, Start, subp2p, vcur, Volume, Warned

Global

Enumerations
MixedWeightOptions	Tags weighting options.	EqualInitialWeights, SimplexWeights, FixedWeights, MixedWeightOptions
QuasiAlgorithms	Tags for Gradient-based optimization algorithms.	USEBHHH, USEBFGS, USEDFP, USESTEEP, USENEWTON, QuasiAlgorithms

BlackBox

BlackBox :: BlackBox ( L )

Create a blackbox objective.

Parameters:

L	string, a label for the problem.

CES

A

decl A [public]

alphas

decl alphas [public]

CES

CES :: CES ( L , alphas , elast , A , labels )

Create a constant elasticity of subsitution objective.

Parameters:

L	label
alphas	CV-compatiable vector of weights/shares. Default is <0.5;0.5>
elast	elasticity parameter ε. The exponent on each input is computed as (elast-1)/elast. Default is -2.0
A	constant term (TFP). Default is 1.0
labels	0 or array of labels for arguments. $$A\left( {\sum}_{i=1}^N \alpha_ix_i^s \right)^{1/s}$$ $$s \equiv {\epsilon-1 \over \epsilon}.$$

Comments:

Sets the parameter vector $x$ equal to StDeviations of the same dimension as alphas. This means it will enforce positive values by transformations. If gradients needed to be computed in terms of the structural input values, then toggle the parameter constraints for the objective:

    U = new CES("U");
    U -> ToggleParameterConstraint();

elast

decl elast [public]

vfunc

virtual CES :: vfunc ( )

Compute CES. $A \left[{\sum} \alpha_i x_i^\sigma\right ]^{1/\sigma}.$ $sigma \equiv (\eta-1)/\eta$

x

decl x [public]

xpon

decl xpon [public]

CobbDouglas

A

decl A [public]

alphas

decl alphas [public]

AnalyticGradient

virtual CobbDouglas :: AnalyticGradient ( )

CobbDouglas

CobbDouglas :: CobbDouglas ( L , alphas , A , labels )

Create a CobbDouglas objective.

Parameters:

label

alphas

CV-compatiable vector of exponents.

labels

0 or array of labels for arguments.

Sets the parameter vector as $x$ equal to StDeviations of the same dimension as alphas

$$A{\prod}_{i=1}^N x_i^{\alpha_i}$$

vfunc

virtual CobbDouglas :: vfunc ( )

$A{\prod}_{i=1}^N x_i^{\alpha_i}$.

x

decl x [public]

Constrained

CheckPoint

virtual Constrained :: CheckPoint ( f , saving )

Store current state of a Constrained Objective (checkpoint to disk).

Parameters:

saving

Comments:

the value of vcur.F is written to the file but ignored by Load().

Constrained

Constrained :: Constrained ( L , ELorN , IELorN )

Create a constrained objective.

Parameters:

L	label
ELorN	Number of equality constraints (integer, array of strings, string list of names)
IELorN	Number of inequality constraints

equality

virtual Constrained :: equality ( )

Default Equality Constraints.

Returns:: empty vector

funclist

Constrained :: funclist ( Fmat , jake )

Compute the objective and constraints at multiple points.

Parameters:

Fmat,

N_f×J matrix of column vectors to evaluate at

Returns:

J, the number of function evaluations

Gradient

virtual Constrained :: Gradient ( extcall )

Compute the ∇f(), objective's gradient at the current vector.

inequality

virtual Constrained :: inequality ( )

Default InEquality Constraints.

Returns:: empty vector

Jacobian

virtual Constrained :: Jacobian ( )

Compute the Jg(), vector version of the Lagrangian's Jacobian at the current vector.

Lagrangian

Constrained :: Lagrangian ( F )

Decode the input, return the whole vector, inequality and equality constraints, if any.

Parameters:

F	vector of free parameters.

Returns:

Array, {ƒ,g,h};

Merit

virtual Constrained :: Merit ( F )

CPoint

Vec

CPoint :: Vec ( )

DataObjective

data

const decl data [public]

path or prediction object.

DataObjective

DataObjective :: DataObjective ( L , data , ... )

An objective based on an economic model with data and (possibly) a nested solution method.

Parameters:

L	string, label
data	an object that includes member FN and method EconometricObjective. Typically, OutcomeDataSet or PredictionDataSet object (including the possibility of the derived OutcomeDataSet and PredictionDataSet variety).
...	Parameters and arrays of Parameters to optimize over.

Comments:

NvfuncTerms is set to data.FN, the total number of paths in the panel

SetStage

DataObjective :: SetStage ( stage )

Set the estimation stage.

Parameters:

stage

Zero, One or Two

Stage Zero:

Parameters on the uplist are set fixed.

Parameters on the tp list can vary .

Bellman iteration method is set to DoNotIterate

Stage One:

uplist variable

tplist fixed

Bellman DoNotIterate = FALSE

Object is reset

Stage Two:

Both parameter lists vary, method iterates

See also:

FPanel::method, PathPrediction::method, ResetMax, Method::DoNotIterate, DoNotVary

stage

decl stage [public]

estimation stage.

tplist

decl tplist [public]

transition parameters .

TwoStage

DataObjective :: TwoStage ( intplist , inuplist )

Specify the objective as two stage.

Parameters:

tplist	a Parameter or a list of parameters that affect the transition only. These will be estimated at Stage 0 or Stage 2
uplist	a Parameter list of parameters that affect Utility only. These will be estimated at Stage 1 or Stage 2.

Comments:

These lists of parameters are added to the overall parameter list so should not be added separately. Parameters that were added when the data objective was created will not be affected by this. They should be toggled separately by the user's code.

uplist

decl uplist [public]

utility parameters .

vfunc

virtual DataObjective :: vfunc ( subp )

Calls and returns data->EconometricObjective() or uses SemiClosedForm() if in Gradient calculation.

Equilibrium

aggF

const decl aggF [public]

list of aggregate prod. function BlackBox.

deprec

const decl deprec [public]

0.0 or vector of quantity depreciations.

Equilibrium

Equilibrium :: Equilibrium ( L , T , P )

Create a system of aggregate equilibrium equations.

Parameters:

L label

T number of periods of static equilibria.
1 [default]

an array of parameters or a parameter block of prices $p$
0 [default] a vector of StDeviations is created of the same length as the parameter list of aggF

This should be called in the creator function of a derived equilibrium class after the following constants have been initialized inside the creator:

aggF : must contain an Objective for the aggregate production function $F(X^d)$. THe parameters of aggF are $X^d$. They are set as DoNotConstrain so that the gradient is correct (not transformed).
stnpred : must contain a PathPrediction object that has a nested solution algorithm. It should typically be set to make predictions from the Ergodic distributions.
Qcols vector of indices into the prediction matrix that correspond to $X^s$. If equal to 0 the first $N$ columns will be selected in order.
deprec CV-compatible vector of input depreciations $\Delta$
0 [default] no depreciation term

foc

decl foc [public]

computed system.

Print

Equilibrium :: Print ( orig , fn , toscreen )

Prints a message and details about the objective.

Parameters:

orig	string, origin of the print call
fn	integer, no print to file. file, prints to file as well as screen
toscreen	TRUE: print full report to screen as well FALSE: only print orig to screen

Q

decl Q [public]

aggregate Qs from prediction.

Qcols

const decl Qcols [public]

columns of pred that match up to inputs.

stnpred

const decl stnpred [public]

ergodic distribution PathPrediction.

T

const decl T [public]

number of periods in equilibrium.

vfunc

virtual Equilibrium :: vfunc ( )

Built-in system of equations for Equilibrium models. Typically the user's derived class does not need to replace this if takes the form below for their model:

Compute prediction using stnpred (which should re-solve DP model)
Get X^s (stored in Q) using GetFlat(DoAll,Qcols)
Encode X^s as the values of the parameter of aggF
Compute and return the equilibrium FOC conditions $$\nabla F(X^s)' - \Delta - P.$$

GHKMNP

Enumerations
Anonymous enum 1		identity, onlydiag, lowertriangle, SigmaOptions

ghk

const decl ghk [public]

GHKMNP

GHKMNP :: GHKMNP ( L , fn , Yname , Xnames , R , iSigma )

GHK based objective for MNP log likelihood.

Parameters:

Yname

Xnames

iSigma

SetGHK

sigfree

const decl sigfree [public]

SigLT

const decl SigLT [public]

vfunc

GHKMNP :: vfunc ( )

Compute and return the vector of log-likelihoods at the current parameters.

GQMNP

GQMNP :: GQMNP ( L , fn , Yname , Xnames , Npts )

Gauss-Hermite based objective for MNP log likelihood.

Parameters:

Npts

Yname

Xnames

Npts

const decl Npts [public]

# of nodes in quadrature

vfunc

GQMNP :: vfunc ( )

Compute and return the vector of log-likelihoods at the current parameters.

MixPoint

aggregate

MixPoint :: aggregate ( outV , v )

Mixture

CheckMax

virtual Mixture :: CheckMax ( fn )

If scur.v > maxpt.v call Save and update maxpt.

Parameters:

file name

Returns:

TRUE if maxval was updated
FALSE otherwise.

D

const decl D [public]

# observed types, environment

Decode

virtual Mixture :: Decode ( F )

Decode and return the parameter array.

Parameters:

F	vector of free parameters.

DK

const decl DK [public]

D x K

dkfree

decl dkfree [public]

DKL

const decl DKL [public]

Dvar

const decl Dvar [public]

Discrete environment-type var

Encode

virtual Mixture :: Encode ( inX )

FinL

decl FinL [public]

Flabels

decl Flabels [public]

fobj

Mixture :: fobj ( F , extcall )

Decode the input, compute the objective, check the maximum.

Parameters:

F	vector of free parameters.

Returns:

mcur.v, f(ψ)

funclist

virtual Mixture :: funclist ( Fmat , aFvec )

Compute the objective at multiple points.

Parameters:

Fmat,	N_f×J matrix of column vectors to evaluate at
aFvec,	address of a ?×J matrix to return `f()` in.
afvec,	0 or address to return aggregated values in.

Returns:

J, the number of function evaluations

Gradient

Compute the ∇f(), objective's gradient at the current vector.

Returns:: scur.G

Included

decl Included [public]

matrix of indicators of valid d,k combos

IncludedDK

Mixture :: IncludedDK ( mDK )

Indicate with (d,k) combinations should be computed.

Parameters:

mDK	D×K matrix of 0s/1s

Comments:

default is ones(D,K)

Jacobian

virtual Mixture :: Jacobian ( )

Compute the Jg(), vector version of the objective's Jacobian at the current vector.

Returns:: Jg(ψ)

lfree

decl lfree [public]

mcur

const decl mcur [public]

Mixture

Mixture :: Mixture ( L , Dvar , Kvar , MixType , ... )

Create a mixture objective.

Parameters:

L	string, a label for the problem.
Dvar,	integer>0, the number of environments Discrete variable that codes the environment
Kvar,	integer>0, the number of types Discrete variable that codes the types
MixType,	MixedWeightOptions
...	D×K matrix or D-array of values No optional argument means Lambda[d] = new Simplex("L"+sprint("%2u",d),K), for d = 0, … D‾ A single argument, then it must be a D×K matrix Lambda[d] = new Simplex("L"+sprint("%2u",d),va[0][d][]), for d = 0, … D‾ D arguments, then each must be a ParameterBlock of size K and Lambda[d] = va[d] for d = 0, … D‾

nfree

decl nfree [public]

vfunc

vobj

virtual Mixture :: vobj ( F )

Decode the input, return the whole vector.

Parameters:

F	vector of free parameters.

WDecode

Mixture :: WDecode ( WF )

WEncode

Mixture :: WEncode ( inW )

Wfunclist

WStart

decl WStart [public]

MLogit

MLogit :: MLogit ( L , fn , Yname , Xnames )

Create a multinomial logit model.

Parameters:

L	label
fn	string, a file to load the data from using Ox `Database.Load()`
Yname	string, name or label of the column in the file that contains `Y` Yname can contain any integers. MNP will translate the unique sorted values into 0...Jvals-1
Xnames	a string of the form var1 var2 ... varN

vfunc

MLogit :: vfunc ( )

Return the log-likelihood vector.

MultiNomialChoice

betas

const decl betas [public]

Array of J-1 parameter blocks, one for each equation except Y=0.

D

decl D [public]

matrix of XB indices.

Estimate

indY

const decl indY [public]

N x J matrix of permutations
the first column is index of Y
second column is first non-Y choice
third column is second non-Y choice
etc.

J

const decl J [public]

J : number of options

Jvals

const decl Jvals [public]

integer codes of choices

lk

decl lk [public]

vector or matrix of likelihods.

MultiNomialChoice

MultiNomialChoice :: MultiNomialChoice ( L , fn , Yname , Xnames )

Create a new Multinomial Choice Model model.

Parameters:

L	label
fn	string, a file to load the data from using Ox `Database.Load()`
Yname	string, name or label of the column in the file that contains `Y` Yname can contain any integers. MNP will translate the unique sorted values into 0...Jvals-1
Xnames	a string of the form var1 var2 ... varN

Comments:

Observations with any missing data are deleted.
A constant column is appended at the end of the X matrix as in Stata.
Summary statistics are reported.
Written with help from Stefan Fassler and Tanvir Ahmed Khan Tanu

namearray

const decl namearray [public]

Labels for variables

NN

const decl NN [public]

indexing vector

nX

const decl nX [public]

K : number of X variables

SetD

MultiNomialChoice :: SetD ( )

Update $XB$ matrix of indices by observation and option.

X

const decl X [public]

X : matrix of exog variables

Y

const decl Y [public]

Y : discrete endog vector

NoObjective

model

decl model [public]

modelmethod

decl modelmethod [public]

NoObjective

NoObjective :: NoObjective ( model )

A wrapper that acts like an objective but just calls a model's Solve method and returns 1.0.

Parameters:

model Object with a method named Solve() or a member named method with a method named Solve()

v

decl v [public]

vfunc

NoObjective :: vfunc ( subp )

Objective

AggSubProbMat

virtual Objective :: AggSubProbMat ( submat )

Block

CheckMax

virtual Objective :: CheckMax ( fn )

If vcur.v > maxpt.v call Save() and update maxpt.

Parameters:

file name

Returns:

TRUE if maxval was updated
FALSE otherwise.

See also:

CheckMaxV

Objective :: CheckMaxV ( v , fn )

Compare v to maxpt.

Parameters:

v	value of objective
fn	file name to print to Output is determined by Volume

CheckPoint

virtual Objective :: CheckPoint ( f , saving )

Load or save state of the problem to a file.

Parameters:

f	file object
saving	TRUE: save status to the file FALSE: load from the file and close it, call Encode

contour

virtual Objective :: contour ( Npts , Xpar , Ypar , lims )

Graph the level curves of the objective in two parameters.

Parameters:

Npts	integer [default=100], number of points to evaluate in each dimensions
Xpar	Parameter for the x axis
Ypar	Parameter for the y axis
lims	2×2 matrix of upper and lower bounds for axes

Decode

virtual Objective :: Decode ( F )

Decode a vector of free variables. Converts an optimized parameter vector into the structural parameter vector. Ensure that each parameter and each parameter block current value is updated.

Parameters:

F,	nfree x 1 vector of optimized parameters. 0 [default], use vcur.F for decode

DoNotConstrain

decl DoNotConstrain [public]

This objective uses raw parameter values.

dStepLinearGap

static const decl dStepLinearGap [public]

Encode

virtual Objective :: Encode ( inX )

Encode vector of structural parameters.

Parameters:

inX	0 [default] get new starting values from the parameters a vector, new starting values.

Comments:

if X is a vector

then a value of .NaN means that parameter should be held fixed at the current value during this cycle of optimization.

If X=0 then the starting values are retrieved as follows:

If this is the first call of Encode() the initial values passed to the parameters when they were created.

If this.X already has elements then this is not the first call to Encode() and starting values will be retrieved from the current values of parameters.

See also:

Recode, nfree, nstruct, FinX

EXT

static const decl EXT [public]

Extension for Load and Save files, =optobj.

Flabels

decl Flabels [public]

array free parameter labels

fname

const decl fname [public]

Name for Load & files

fobj

virtual Objective :: fobj ( F , extcall )

Decode the input, compute the objective, check the maximum.

Parameters:

F	vector of free parameters.
extcall	TRUE [default] go into server mode if MPI server, send Stop if client

FreeStatus

Objective :: FreeStatus ( Store )

Preserve the state of free or fixed status of the parameter vector.

Parameters:

Store

TRUE, store the current status
FALSE restore after a previous store

See also:

ToggleParams, ToggleBlockElements

fshold

decl fshold [public]

holds DoNotVary vector for temp. 1-d solution.

funclist

virtual Objective :: funclist ( Fmat , aFvec , afvec , abest )

Compute the objective at multiple points.

Parameters:

Fmat,	N_f×J matrix of column vectors to evaluate at
aFvec,	address of a ?×J matrix to return `f()` in.
afvec,	0 or address to return aggregated values in (as a 1 × J ROW vector)
abest,	0 or address to return index of best vector The maximum value is also computed and checked.

Returns:

J, the number of function evaluations

Gradient

virtual Objective :: Gradient ( extcall )

Compute the ∇f(), objective's gradient at the current vector.

Parameters:

extcall

[default=TRUE]. if running in parallel Stop message sent out

Compute

$$\nabla f(\psi)$$ Stored in vcur.G

Any derived Objective can replace this virtual method with one that computes analytic gradients or other approaches.

Returns:

vcur.G

Hessian

virtual Objective :: Hessian ( )

Compute the Hf(), Hessian of objective at the current vector.

Returns:: Hg(ψ) in vcur.H

INGRADIENT

decl INGRADIENT [public]

TRUE when computing Gradient, can be used by semi-closed form gradient methods.

Interact

Jacobian

virtual Objective :: Jacobian ( )

Compute the Jg(), vector version of the objective's Jacobian at the current vector.

Returns:: Jg(ψ) in vcur.J

L

const decl L [public]

label.

Load

Objective :: Load ( fname )

Load state of the problem.

Parameters:

fname

string, name of file
0 [default], use the default name, the label with spaces removed
-1, do nothing and return FALSE.

Returns:

TRUE if parameter values were loaded from file
FALSE otherwise.

Comments:

FinX is read in to set DONOTVARY for each parameter. The value of F is ignored by Load. Encode() called by Load.

See also:

Save, L

logf

decl logf [public]

log file

lognm

const decl lognm [public]

name of log file

maxpt

const decl maxpt [public]

Best so far

See also:: CheckMax

MyVersion

static decl MyVersion [public]

Version of niqlow that code is written for.

See also:: SetVersion

newmax

decl newmax [public]

Set in CheckMax(), TRUE if latest check was new max.

nfree

decl nfree [public]

length of F in current cycle.

nstruct

decl nstruct [public]

length of X.

NvfuncTerms

decl NvfuncTerms [public]

Longest vector returned by vfunc(), default=1.

Objective

Objective :: Objective ( L , CreateCur )

Create a new objective.

Parameters:

L	string, label for the objective.
CreateCur	create a new Point in vcur

once

decl once [public]

TRUE if encoded once

p2p

decl p2p [public]

P2P object for using MPI across param vectors

Parameters

virtual Objective :: Parameters ( ... )

Add Parameters to the parameter list to be optimized over.

Parameters:

...	the first or next (and possibly only) parameter to add to the objective array: any argument can send an array which contains only parameters and blocks

See also:

Parameter, Psi

Print

virtual Objective :: Print ( orig , fn , toscreen )

Prints a message and details about the objective.

Parameters:

orig	string, origin of the print call
fn	integer, no print to file. file, prints to file as well as screen
toscreen	TRUE: print full report to screen as well FALSE: only print orig to screen

PsiL

decl PsiL [public]

all parameter labels

PsiType

decl PsiType [public]

parameter class names

Recode

virtual Objective :: Recode ( HardCode )

Reset the free parameter vector and the complete structural parameter vector.

Parameters:

HardCode

TRUE=hard-coded parameter values used.
FALSE [default] Start vector used Must be called anytime a change in constraints is made. So it is called by Encode() and ReInitialize()

ReInitialize

virtual Objective :: ReInitialize ( )

Revert all parameters to their hard-coded initial values. This routine can only be called after Encode() has been called.

All parameters are reset to their hard-coded initial values, stored as ival.

This also sets the values of the vcur vector, including the structural parameters X and free parameters F

ResetMax() is called

The result is as if no optimization has occurred and Encode(0) has just been executed for the first time.

ResetMax

virtual Objective :: ResetMax ( )

Reset the value of the current and maximum objective.

RunSafe

decl RunSafe [public]

TRUE (default): exit if NaNs encountered during iteration
FALSE: exit with IterationFailed

Save

Objective :: Save ( fname )

Store current state (checkpoint to disk).

Parameters:

fname

string, name of file
0 [default] use fname

See also:

Load, EXT

SetAggregation

Objective :: SetAggregation ( AggType )

Set the formula .

Parameters:

AggType

See also:

AggregatorTypes

SetVersion

static Objective :: SetVersion ( v )

Checks the version number you send with the current version of niqlow.

Parameters:

v	integer [default=200]

Start

decl Start [public]

Initial vector after Encode()

subp2p

decl subp2p [public]

P2P object for MPI on a single param vector.

ToggleBlockElements

virtual Objective :: ToggleBlockElements ( pblock , elements )

Toggle DoNotVary for one or more parameters.

Parameters:

pblock	ParameterBlock
elements	vector of indices of block elements to toggle.

ToggleParameterConstraint

Objective :: ToggleParameterConstraint ( )

Toggle the value of DoNotConstrain. If DoNotConstrain then all parameters except Determined parameters are free and unscaled. This can improve convergence near the top of the objective. It also means that the Hessian values are in terms of the economic parameters not the free-to-vary optimization parameters.

ToggleParams

virtual Objective :: ToggleParams ( ... )

Toggle DoNotVary for one or more parameters.

Parameters:

...	Parameters or arrays of parameter To toggle elements of a parameter block ...

See also:

ToggleBlockElements

vcur

const decl vcur [public]

current point.

vfunc

virtual Objective :: vfunc ( subp )

Built in objective, f(ψ). Prints a warning once and then returns 0.

vobj

virtual Objective :: vobj ( F )

Decode the input, return the whole vector.

Parameters:

F	vector of free parameters.

Returns:

vcur.V

Volume

decl Volume [public]

Output Level.

Warned

static decl Warned [public]

. internal

OneDimSystem

CheckMax

OneDimSystem :: CheckMax ( fn )

This is misnamed. It checks whether the 1-D system is closer to 0.

Parameters:

file name

ipar

decl ipar [public]

isys

decl isys [public]

msys

const decl msys [public]

OneDimSystem

OneDimSystem :: OneDimSystem ( L , msys )

Create a new non-linear equation.

Parameters:

L	label for the system.
msys	System object that is the mother system 0 [default] no mother system A system of this type can be solved using OneDimSolve

ResetMax

OneDimSystem :: ResetMax ( )

Reset the value of the system. reversed sign from others.

SetOneDim

OneDimSystem :: SetOneDim ( insys , inpar )

Parameters:

insys

inpar

vfunc

virtual OneDimSystem :: vfunc ( )

Separable

C

decl C [public]

Number of common parameters

CDNV

decl CDNV [public]

CheckMax

virtual Separable :: CheckMax ( fn )

If scur.v > maxpt.v call Save and update maxpt.

Parameters:

file name

Returns:

TRUE if maxval was updated
FALSE otherwise.

ComInd

decl ComInd [public]

Vector of indices into Psi of common pars

Common

virtual Separable :: Common ( ... )

Add parameters to the objective that will have a single value across types/sub-problems.

Parameters:

...	Parameter(s) and/or arrays of Parameters to add to the objective

Comments:

On any call to vfunc() common parameters will have the same value for each k.

Decode

virtual Separable :: Decode ( F )

Decode and return the structural parameter matrix.

Parameters:

F	vector of free parameters.

Returns:

Encode

virtual Separable :: Encode ( X , CallBase )

Encode matrix of structural parameters.

Parameters:

X	0 get starting values from current X vector N×1, common starting values across types matrix N×K, starting values for all types. Only first column value used for common types
CallBase	TRUE, simply call Encode(), used in parallel processing. FALSE (default), proceed with separable encoding

Flabels

decl Flabels [public]

fobj

virtual Separable :: fobj ( F , extcall )

Decode the input, compute the objective, check the maximum.

Parameters:

F	vector of free parameters.

Returns:

f(ψ)

funclist

virtual Separable :: funclist ( Fmat , aFvec , afvec )

Compute the objective at multiple points.

Parameters:

Fmat,	N_f×J matrix of column vectors to evaluate at
aFvec,	address of a ?×J matrix to return `f()` in.
afvec,	0 or address to return aggregated values in.

Returns:

J, the number of function evaluations

Gradient

virtual Separable :: Gradient ( extcall )

Compute the ∇f(), objective's gradient at the current vector.

Returns:: scur.G

Included

decl Included [public]

Jacobian

virtual Separable :: Jacobian ( )

Compute the Jg(), vector version of the objective's Jacobian at the current vector. Returns Jg(ψ) in cur.J.

K

const decl K [public]

# unobserved types, sub-problems

kfree

decl kfree [public]

K Vector of free specific parameters

KL

const decl KL [public]

labels for types / sub problems

kNvf

decl kNvf [public]

Kvar

const decl Kvar [public]

Discrete sub-problem var

nfree

decl nfree [public]

Total free parameters

NvfuncTerms

decl NvfuncTerms [public]

Longest vector returned by vfunc(), default=1.

Print

virtual Separable :: Print ( orig , fn , toscreen )

scur

const decl scur [public]

Separable

Separable :: Separable ( L , Kvar )

Create a separable objective.

Parameters:

L	string, a label for the problem.
Kvar,	integer>0, the number of sub objectives Discrete variable that codes the problem vector or ParameterBlock, weights on values of k

vfunc

vobj

virtual Separable :: vobj ( F )

Decode the input, return the whole vector.

Parameters:

F	vector of free parameters.

Returns:

svfunc()

System

EndOneDim

System :: EndOneDim ( )

eq1

decl eq1 [public]

1 equation of system.

equations

virtual System :: equations ( )

Default system of equations: vfunc().

normexp

decl normexp [public]

SetOneDim

virtual System :: SetOneDim ( insys , inpar )

Parameters:

insys	if eq1 an object then the equation index to select [default=0]
inpar	if eq1 an object the parameter or parameter index to vary.

System

System :: System ( L , LorN , incur )

Create a new system of equations object.

Parameters:

L	label for the system.
LorN	The size of the system indicated in 1 of 3 ways: integer [default = 1], N the size of the system array of length N, where each element is a string, the label for the equation. string with N space-separate labels, the labels of the equations parsed by varlist.

Example:

Three ways to define a 3x3 system of equations:

 mysys = new System(3);
 mysys = new System({"A","B","C"});
 mysys = new System("Eq0 Eq2 Eq2");

See also:

NvfuncTerms, Equations

UnConstrained

Gradient

virtual UnConstrained :: Gradient ( extcall )

Compute the ∇f(), objective's gradient at the current vector.
Compute

UnConstrained

UnConstrained :: UnConstrained ( L )

Create an unconstrained objective.

Parameters:

label